AEA(Aria Everyday Activities):只用 speech + sound 事件做 goal / sub-goal 标注 — 统计 + demo(2026-10-27)

标注来源:savvy-bench speech = SAVVY-Bench(Chen et al., NeurIPS 2025)出题用的句子,带说话人(is_me_speak)、partner 方向/距离(speech_gt_keyframes);savvy-bench sound = 同上的声音事件;ours-legacy sound = 旧项目自动声音标签,未核验;goal / sub-goal = Gemini 3.1 Pro 草稿(Method E 类比:official script activity type 当「官方 step」,sub-goal 只引用事件 id,时间由代码从事件复制),未人工核验。AEA 官方没有 narration / recipe / protocol;本页不用 legacy narration。

1. 有多少可用

recordings with SAVVY window: 62; total speech annos 661 (wearer 127 / partner 534), savvy sound 16, legacy sound in window 864
recordings with >= 6 speech annos: 51; >= 10: 34; >= 15: 11
median speech annos per recording: 10.0; median window 138 s; speech annos per minute (median): 4.5
pairs where both sides annotated: 30; local VRS: 8; savvy mp4: 62

SAVVY 只标了 script 2/3 的双人录像(62 条 = 30 对 + loc2_script3_seq3 拆成 4 段)。speech 标注 = 出题句,不是全量对话;legacy sound 为未核验的弱证据。

62 条逐条统计(按 speech 标注数排序)

recording	window s	speech (wearer/partner)	sound savvy	sound legacy	savvy mp4	local VRS	partner side annotated
loc1_script2_seq7_rec2 example	270	28 (0/28)	0	7	Y	–	Y
loc1_script2_seq7_rec1 example	265	26 (8/18)	0	4	Y	–	Y
loc3_script2_seq7_rec1 example	100	25 (12/13)	0	1	Y	–	Y
loc3_script2_seq7_rec2	99	25 (6/19)	0	1	Y	–	Y
loc2_script2_seq5_rec1 example	255	24 (0/24)	0	14	Y	–	Y
loc3_script3_seq2_rec1 example	169	22 (0/22)	0	3	Y	Y	Y
loc2_script3_seq4_rec1 example	250	20 (1/19)	0	22	Y	–	Y
loc2_script2_seq2_rec1 example	106	18 (4/14)	0	5	Y	–	Y
loc3_script2_seq3_rec2 example	188	18 (4/14)	1	13	Y	–	Y
loc3_script2_seq3_rec1 example	189	16 (2/14)	2	12	Y	Y	Y
loc2_script2_seq1_rec1 example	134	15 (3/12)	1	9	Y	–	Y
loc1_script2_seq3_rec1 example	196	14 (1/13)	1	13	Y	–	Y
loc2_script2_seq4_rec1	101	14 (4/10)	0	4	Y	–	Y
loc2_script2_seq1_rec2	133	13 (2/11)	1	8	Y	–	Y
loc2_script3_seq31_rec1	160	13 (5/8)	0	24	Y	–	Y
loc3_script3_seq4_rec1	281	13 (4/9)	0	43	Y	–	Y
loc2_script2_seq8_rec2	94	12 (1/11)	1	1	Y	–	Y
loc2_script3_seq33_rec2	156	12 (3/9)	0	6	Y	–	Y
loc3_script2_seq1_rec2	113	12 (0/12)	0	12	Y	–	Y
loc3_script3_seq2_rec2	173	12 (0/12)	0	5	Y	–	Y
loc2_script2_seq4_rec2 example	99	11 (2/9)	0	8	Y	Y	Y
loc2_script3_seq4_rec2 example	254	11 (8/3)	1	21	Y	Y	Y
loc1_script2_seq1_rec1	120	10 (2/8)	2	3	Y	–	Y
loc1_script2_seq3_rec2	195	10 (5/5)	0	24	Y	–	Y
loc2_script3_seq2_rec1	82	10 (0/10)	0	1	Y	–	Y
loc2_script3_seq2_rec2	82	10 (0/10)	0	4	Y	–	Y
loc2_script3_seq32_rec1	170	10 (2/8)	0	19	Y	–	Y
loc2_script3_seq32_rec2	170	10 (3/7)	0	15	Y	–	Y
loc2_script3_seq33_rec1	156	10 (5/5)	0	15	Y	–	Y
loc2_script3_seq34_rec1	180	10 (6/4)	0	10	Y	–	Y
loc2_script3_seq5_rec2	229	10 (0/10)	0	21	Y	–	Y
loc3_script2_seq1_rec1	114	10 (2/8)	0	5	Y	–	Y
loc3_script2_seq5_rec1	131	10 (0/10)	0	26	Y	Y	Y
loc3_script3_seq4_rec2	278	10 (1/9)	0	23	Y	–	Y
loc1_script2_seq1_rec2	119	9 (0/9)	2	3	Y	–	Y
loc1_script2_seq4_rec1	213	9 (0/9)	0	20	Y	–	Y
loc1_script2_seq4_rec2	209	9 (0/9)	0	15	Y	–	Y
loc2_script2_seq3_rec2	79	9 (0/9)	0	2	Y	–	Y
loc2_script2_seq8_rec1	94	9 (2/7)	0	2	Y	–	Y
loc2_script3_seq31_rec2	159	9 (2/7)	0	20	Y	–	Y
loc1_script2_seq6_rec2	141	8 (1/7)	1	24	Y	–	Y
loc1_script2_seq8_rec1	94	8 (0/8)	1	3	Y	–	Y
loc2_script2_seq2_rec2	105	8 (2/6)	0	2	Y	–	Y
loc2_script3_seq34_rec2	180	8 (1/7)	0	47	Y	–	Y
loc3_script2_seq4_rec1	116	7 (0/7)	0	10	Y	–	Y
loc3_script3_seq1_rec2	64	7 (3/4)	0	3	Y	–	Y
loc1_script2_seq8_rec2	100	6 (1/5)	1	8	Y	–	Y
loc2_script2_seq3_rec1	80	6 (3/3)	0	2	Y	Y	Y
loc2_script2_seq5_rec2	256	6 (1/5)	0	25	Y	–	Y
loc2_script3_seq1_rec1	33	6 (4/2)	0	3	Y	–	Y
loc2_script3_seq1_rec2	36	6 (2/4)	0	2	Y	–	Y
loc1_script2_seq6_rec1	125	5 (3/2)	0	9	Y	–	Y
loc2_script2_seq6_rec2	65	5 (0/5)	0	9	Y	–	Y
loc2_script3_seq5_rec1	239	5 (3/2)	0	18	Y	–	Y
loc3_script3_seq5_rec2	190	5 (1/4)	1	16	Y	–	Y
loc2_script2_seq6_rec1	62	4 (0/4)	0	5	Y	–	Y
loc3_script2_seq4_rec2	114	4 (0/4)	0	6	Y	–	Y
loc3_script2_seq5_rec2	134	4 (0/4)	0	23	Y	–	Y
loc3_script3_seq1_rec1	63	4 (2/2)	0	7	Y	Y	Y
loc3_script3_seq5_rec1	191	1 (0/1)	0	22	Y	Y	Y
loc2_script3_seq3_rec1	666	0 (0/0)	0	68	Y	–	–
loc2_script3_seq3_rec2	666	0 (0/0)	0	88	Y	–	–

2. Demo:13 个 example(含 3 对同场景两侧录像)

loc3_script2_seq3_rec1[making coffee] Make a cup of coffee for the partner.
189 s · 16 speech · 4 sub-goals
audio_helpsspeech_onlyaudio_helpsaudio_helps

loc3_script2_seq3_rec2[making coffee] Make French press coffee for two
188 s · 18 speech · 4 sub-goals
audio_helpsspeech_onlyaudio_helpsaudio_helps

loc2_script3_seq4_rec2[dining] Set the table and eat dinner together.
254 s · 11 speech · 2 sub-goals
speech_onlyspeech_only

loc2_script3_seq4_rec1[dining] Set the table and eat dinner together.
250 s · 20 speech · 2 sub-goals
speech_onlyspeech_only

loc1_script2_seq7_rec1[multi-person indoor activity (games, conversation)] Select and watch a movie together on the TV.
265 s · 26 speech · 3 sub-goals
speech_onlyspeech_onlyspeech_only

loc1_script2_seq7_rec2[multi-person indoor activity (games, conversation)] Watch a movie together and converse
270 s · 28 speech · 3 sub-goals
speech_onlyspeech_onlyspeech_only

loc2_script2_seq4_rec2[preparing snacks] Prepare popcorn and coffee to enjoy with the guest.
99 s · 11 speech · 3 sub-goals
speech_onlyspeech_onlyspeech_only

loc2_script2_seq5_rec1[multi-person indoor activity (games, conversation)] Play a game of Scrabble and eat snacks with the partner.
255 s · 24 speech · 3 sub-goals
speech_onlyspeech_onlyspeech_only

loc2_script2_seq2_rec1[touring the room] Take a guided tour of the host's home
106 s · 18 speech · 4 sub-goals
speech_onlyspeech_onlyspeech_onlyspeech_only

loc2_script2_seq1_rec1[multi-person indoor activity (games, conversation)] Enter the host's place, get a drink, and catch up.
134 s · 15 speech · 3 sub-goals
audio_helpsaudio_helpsspeech_only

loc1_script2_seq3_rec1[making coffee] Make and serve coffee for the guest
196 s · 14 speech · 4 sub-goals
speech_onlyaudio_helpsspeech_onlyaudio_helps

loc3_script2_seq7_rec1[multi-person indoor activity (games, conversation)] Watch a stock market video and discuss investments with the partner.
100 s · 25 speech · 2 sub-goals
speech_onlyspeech_only

loc3_script3_seq2_rec1[multi-person indoor activity (games, conversation)] Sit down and converse with the partner about recent trips and experiences.
169 s · 22 speech · 4 sub-goals
speech_onlyspeech_onlyspeech_onlyspeech_only

Making coffee together (pair, both views)

loc3_script2_seq3_rec1

script 2 · window 189 s · speech annos 16(wearer 2 / partner 14)· savvy sound 2 · legacy sound 12 · partner recording: loc3_script2_seq3_rec2

video: SAVVY Video/loc3_script2_seq3_rec1.mp4(20 fps 去畸变小视频,音轨 = mic 5+6 stereo;时间 0 = device 460.20 s)

上条:sub-goal 区段(点击跳转);下条:事件刻度,绿 = wearer 说话,橙 = partner 说话,蓝 = SAVVY sound,灰 = legacy sound

Goal / sub-goal 草稿(Gemini 3.1 Pro,candidate (Gemini draft, not human verified))

activity_type(官方列表选一):making coffee
overall_goal:Make a cup of coffee for the partner.
roles:wearer = host, partner = guest
unrelated events:S003, S011 · validation errors: 0

#	sub-goal	clip time	initiated_by	completion	audio_necessity	events	rationale
g1	Fill the kettle with water and ignite the gas stove.	0:01.3–0:53.9	partner_request	mixed	audio_helps	S000 S001 X000 X001 S002 X002 A000 A001 X003	The partner requests coffee, prompting the wearer to fill a container with tap water and ignite the gas burner to heat it.
g2	Locate the coffee filters and examine the coffee maker.	0:59.7–1:21.5	wearer_self	speech	speech_only	X004 S004 X005 S005 S006 S007	The wearer searches for coffee filters while the partner finds a fancy coffee maker and comments on not knowing how it works.
g3	Wait for the water to boil and pour it.	1:33.0–2:06.0	joint_decision	mixed	audio_helps	X006 S008 X007 X008 S009 S010	They wait for the water to finish boiling, as evidenced by the bubbling sounds, and then discuss pouring it.
g4	Operate the coffee maker, pour the brewed coffee, and evaluate it.	2:13.2–3:01.3	joint_decision	mixed	audio_helps	S012 X009 S013 X010 S014 X011 S015	The partner instructs the wearer on how to press the button on the coffee maker, followed by pouring sounds and comments on the coffee's strength.

事件时间线(只用 speech + sound,不用 narration)

id	clip time	kind	speaker / object	text	partner direction · distance	source
S000	0:01.3–0:02.9	speech	partner	I would love a cup of coffee. topic: expressing desire for coffee	front-right · 1.36 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S001	0:11.9–0:13.0	speech	partner	So you have to put water in this? topic: asking about adding water to coffee maker	front · 1.08 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X000	0:23.6–0:30.3	sound	sink	water filling	–	ours-legacy sound (unverified, model/hand made)
X001	0:30.5–0:31.9	sound	tap	tap water running	–	ours-legacy sound (unverified, model/hand made)
S002	0:36.2–0:37.9	speech	wearer	Do we get a little more water in here maybe? topic: asking about adding more water	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X002	0:38.1–0:41.5	sound	tap	tap water running	–	ours-legacy sound (unverified, model/hand made)
A000	0:49.8–0:52.8	sound	hob	ticking	front-left · 0.35 m	savvy-bench sound (QA-selected event)
A001	0:52.3–0:53.0	sound	hob	flame	front-left · 0.39 m	savvy-bench sound (QA-selected event)
X003	0:53.4–0:53.9	sound	gas stove	gas burner igniting	–	ours-legacy sound (unverified, model/hand made)
S003	0:54.9–0:57.6	speech	partner	We'll just do it for a little while. We don't have to hear it all yet. topic: suggesting running process briefly	front-left · 1.16 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X004	0:59.7–1:00.3	sound	package	opening package	–	ours-legacy sound (unverified, model/hand made)
S004	1:02.6–1:04.7	speech	wearer	We have some coffee filters or something like that maybe. topic: inquiring about coffee filters	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X005	1:04.5–1:05.2	sound	cabinet	opening cabinet	–	ours-legacy sound (unverified, model/hand made)
S005	1:05.0–1:06.0	speech	partner	We have this thing. topic: stating possession of filter alternative	front-left · 1.97 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S006	1:13.3–1:14.7	speech	partner	I don't know how this works. topic: admitting not knowing how item works	front-left · 1.54 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S007	1:20.9–1:21.5	speech	partner	It's fancy. topic: describing item as fancy	front · 0.74 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X006	1:33.0–1:33.2	sound	water	boiling	–	ours-legacy sound (unverified, model/hand made)
S008	1:33.9–1:35.8	speech	partner	We'll just wait like a couple of seconds. topic: suggesting waiting briefly	front-left · 0.76 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X007	1:46.4–1:57.3	sound	water	water bubbling	–	ours-legacy sound (unverified, model/hand made)
X008	1:46.4–1:57.3	sound	water	water pouring	–	ours-legacy sound (unverified, model/hand made)
S009	2:01.4–2:04.5	speech	partner	I believe we have to pour it into mugs. topic: stating belief coffee needs pouring into mugs	front-left · 0.73 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S010	2:04.5–2:06.0	speech	partner	We could just do it in. Yup. topic: agreeing to pour directly into mugs	front-left · 0.78 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S011	2:09.8–2:11.6	speech	partner	And then pretend to take a sip. topic: suggesting pretending to sip	front-left · 0.67 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S012	2:13.2–2:14.7	speech	partner	That's so weak. topic: commenting on coffee weakness	front-left · 0.58 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X009	2:27.7–2:28.4	sound	scissors	Cut	–	ours-legacy sound (unverified, model/hand made)
S013	2:28.0–2:29.3	speech	partner	Press this button back here. topic: instructing to press specific button	front-left · 0.57 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X010	2:36.7–2:49.3	sound	coffee	pour	–	ours-legacy sound (unverified, model/hand made)
S014	2:45.5–2:48.3	speech	partner	You don't have to move it like for the French press by the way. topic: instructing not to move French press while pressing	front-left · 0.34 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X011	2:53.5–2:54.9	sound	drinking	gulp	–	ours-legacy sound (unverified, model/hand made)
S015	3:00.2–3:01.3	speech	partner	You only do it the one time. topic: stating action is done only once	front-right · 0.78 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)

loc3_script2_seq3_rec2

script 2 · window 188 s · speech annos 18(wearer 4 / partner 14)· savvy sound 1 · legacy sound 13 · partner recording: loc3_script2_seq3_rec1

video: SAVVY Video/loc3_script2_seq3_rec2.mp4(20 fps 去畸变小视频,音轨 = mic 5+6 stereo;时间 0 = device 585.21 s)

上条:sub-goal 区段(点击跳转);下条:事件刻度,绿 = wearer 说话,橙 = partner 说话,蓝 = SAVVY sound,灰 = legacy sound

Goal / sub-goal 草稿(Gemini 3.1 Pro,candidate (Gemini draft, not human verified))

activity_type(官方列表选一):making coffee
overall_goal:Make French press coffee for two
roles:wearer = unknown, partner = unknown
unrelated events:X012 · validation errors: 0

#	sub-goal	clip time	initiated_by	completion	audio_necessity	events	rationale
g1	Fill the kettle with water and turn on the stove	0:17.3–0:52.4	partner_request	mixed	audio_helps	S000 X000 X001 S001 X002 X004	The partner suggests getting water from the sink, followed by sounds of water filling the kettle and the gas stove turning on.
g2	Examine the coffee maker and add coffee grounds	0:40.0–1:29.4	wearer_self	speech	speech_only	X003 X005 S002 S003 S004 S005	The wearer examines the fancy coffee maker and the partner suggests putting the coffee in the bottom.
g3	Check the boiling water and pour it into the coffee maker	1:31.7–2:17.8	partner_request	mixed	audio_helps	S006 A000 X006 S007 S008 S009 X007 X008 X009 S010	The partner notes the water is boiling, asks to turn it off, and directs pouring the water into the coffee maker.
g4	Press the coffee, pour it into cups, and drink	2:19.1–2:54.2	partner_request	mixed	audio_helps	S011 S012 S013 S014 S015 X010 S016 S017 X011	The partner initiates using the French press, asks about cups, pours the coffee, and they toast before drinking.

事件时间线(只用 speech + sound,不用 narration)

id	clip time	kind	speaker / object	text	partner direction · distance	source
S000	0:17.3–0:18.5	speech	partner	Probably want to get it from the sink. topic: suggesting getting water from sink	front · 0.78 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X000	0:22.2–0:30.9	sound	kitchen sink	faucet filling a kettle	–	ours-legacy sound (unverified, model/hand made)
X001	0:29.5–0:32.5	sound	sink	water	–	ours-legacy sound (unverified, model/hand made)
S001	0:35.8–0:37.5	speech	partner	Do we get a little more water in here maybe? topic: asking about adding more water	front · 1.64 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X002	0:38.5–0:38.8	sound	sink	water	–	ours-legacy sound (unverified, model/hand made)
X003	0:40.0–0:46.0	sound	bag	zipping	–	ours-legacy sound (unverified, model/hand made)
X004	0:50.7–0:52.4	sound	gas stove	gas stove	–	ours-legacy sound (unverified, model/hand made)
X005	0:59.5–1:00.8	sound	bag	tearing	–	ours-legacy sound (unverified, model/hand made)
S002	1:04.6–1:05.6	speech	wearer	We have this thing. topic: stating possession of filter alternative	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S003	1:12.9–1:14.3	speech	wearer	I don't know how this works. topic: admitting not knowing how item works	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S004	1:20.5–1:21.1	speech	wearer	It's fancy. topic: describing item as fancy	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S005	1:27.8–1:29.4	speech	partner	Put the coffee in the bottom, right? topic: confirming coffee placement in bottom	front-right · 0.8 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S006	1:31.7–1:33.3	speech	partner	This water nice and boiling. topic: observing water is boiling	front-right · 0.82 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
A000	1:33.6–1:43.6	sound	hob	background boiling	front · 0.81 m	savvy-bench sound (QA-selected event)
X006	1:33.7–1:33.7	sound	kettle	boiling	–	ours-legacy sound (unverified, model/hand made)
S007	1:40.9–1:41.8	speech	partner	Turn that off. topic: instructing to turn off heat	front-right · 0.76 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S008	1:42.9–1:43.9	speech	partner	Pop that open. topic: instructing to open something	front-right · 0.73 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S009	1:44.0–1:45.3	speech	partner	Let's pour this in. topic: suggesting pouring water	front-right · 0.66 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X007	1:45.3–1:56.5	sound	kettle	water pouring	–	ours-legacy sound (unverified, model/hand made)
X008	2:04.6–2:06.8	sound	cabinet	Cabinet opening sound.	–	ours-legacy sound (unverified, model/hand made)
X009	2:09.4–2:09.9	sound	cabinet	Cabinet closing sound.	–	ours-legacy sound (unverified, model/hand made)
S010	2:15.7–2:17.8	speech	partner	The water's not fully hot. topic: stating water wasn't fully hot	front-right · 0.6 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S011	2:19.1–2:20.1	speech	partner	Let's French press it. topic: suggesting using French press method	front-right · 0.6 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S012	2:20.2–2:21.2	speech	partner	Let's get that action. topic: encouraging French press action	front-right · 0.59 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S013	2:29.0–2:29.8	speech	partner	Click that in. topic: instructing clicking something into place	front-right · 0.6 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S014	2:30.8–2:32.0	speech	partner	Alright, yeah. We should be good. topic: confirming readiness	front-right · 0.64 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S015	2:32.7–2:34.6	speech	partner	You want a cup too? Cup for both of us? topic: offering coffee to both participants	front-right · 0.6 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X010	2:35.4–2:47.9	sound	coffee	pouring	–	ours-legacy sound (unverified, model/hand made)
S016	2:45.1–2:47.9	speech	wearer	You don't have to move it like for the French press by the way. topic: instructing not to move French press while pressing	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S017	2:50.7–2:51.6	speech	partner	Bottoms up. topic: offering a toast	front-right · 0.66 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X011	2:51.1–2:54.2	sound	person	drinking	–	ours-legacy sound (unverified, model/hand made)
X012	3:06.5–3:07.8	sound	phone	Typing on phone	–	ours-legacy sound (unverified, model/hand made)

Setting the table and dining (pair, both views)

loc2_script3_seq4_rec2

script 3 · window 254 s · speech annos 11(wearer 8 / partner 3)· savvy sound 1 · legacy sound 21 · partner recording: loc2_script3_seq4_rec1

video: SAVVY Video/loc2_script3_seq4_rec2.mp4(20 fps 去畸变小视频,音轨 = mic 5+6 stereo;时间 0 = device 4550.28 s)

上条:sub-goal 区段(点击跳转);下条:事件刻度,绿 = wearer 说话,橙 = partner 说话,蓝 = SAVVY sound,灰 = legacy sound

Goal / sub-goal 草稿(Gemini 3.1 Pro,candidate (Gemini draft, not human verified))

activity_type(官方列表选一):dining
overall_goal:Set the table and eat dinner together.
roles:wearer = unknown, partner = unknown
unrelated events:S005 · validation errors: 0

#	sub-goal	clip time	initiated_by	completion	audio_necessity	events	rationale
g1	Gather plates, utensils, food, and drinks to set the table.	0:07.1–1:11.2	wearer_self	mixed	speech_only	S000 X000 S001 X001 X002 S002 X003 S003 S004 X004	The wearer initiates getting plates, food, and forks, and asks the partner to grab drinks, which is supported by drawer and table setting sounds.
g2	Eat the meal and discuss the doneness of the meat.	1:45.6–3:58.2	joint_decision	mixed	speech_only	S006 S007 S008 S009 S010 X005 X006 X007 A000 X008 X009 X010 X011 X012 X013 X014 X015 X016 X017 X018 X019	The wearer asks about the meat, leading to a conversation about it being underdone, accompanied by sounds of eating, opening a can, and using cutlery.

事件时间线(只用 speech + sound,不用 narration)

id	clip time	kind	speaker / object	text	partner direction · distance	source
S000	0:07.1–0:08.6	speech	wearer	Let's grab some some plates. topic: suggesting grabbing plates	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X000	0:17.8–0:18.5	sound	drawer	drawer open	–	ours-legacy sound (unverified, model/hand made)
S001	0:26.5–0:27.6	speech	wearer	I'll bring the the food. topic: volunteering to bring the food	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X001	0:27.5–0:28.1	sound	drawer	drawer open	–	ours-legacy sound (unverified, model/hand made)
X002	0:29.2–0:29.4	sound	drawer	Drawer opening or closing	–	ours-legacy sound (unverified, model/hand made)
S002	0:29.7–0:30.2	speech	wearer	Um couple of forks. topic: mentioning needing forks	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X003	0:35.0–0:35.3	sound	drawer	Drawer opening or closing	–	ours-legacy sound (unverified, model/hand made)
S003	0:36.2–0:37.8	speech	wearer	Alright and then if you want to grab a couple of drinks. topic: suggesting grabbing drinks	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S004	0:39.8–0:41.8	speech	partner	Grab some lacrosse for us. topic: requesting grabbing La Croix drinks	back-right · 2.43 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X004	0:59.4–1:11.2	sound		Table setting sounds.	–	ours-legacy sound (unverified, model/hand made)
S005	1:12.5–1:13.4	speech	partner	Do we eat in this one? topic: asking about eating location or room	front-right · 0.84 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S006	1:45.6–1:46.4	speech	wearer	How's the meat. topic: asking about meat quality	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S007	1:47.4–1:48.2	speech	partner	A little underdone. topic: describing meat as slightly underdone	back-left · 5.32 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S008	1:48.7–1:50.1	speech	wearer	Well, that's not terrible. topic: commenting that underdone meat is acceptable	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S009	1:50.2–1:51.3	speech	wearer	It's not ground beef. topic: differentiating meat type from ground beef regarding doneness	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S010	1:53.7–1:55.6	speech	wearer	I mean, I like beef rare. topic: expressing preference for rare beef	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X005	2:17.0–2:26.8	sound	human	eating	–	ours-legacy sound (unverified, model/hand made)
X006	2:32.4–2:33.4	sound	human	eating	–	ours-legacy sound (unverified, model/hand made)
X007	2:37.0–2:39.3	sound	human	food chewing	–	ours-legacy sound (unverified, model/hand made)
A000	2:40.5–2:43.5	sound	dining table	a tab being pulled open	front-right · 0.46 m	savvy-bench sound (QA-selected event)
X008	2:42.6–2:43.9	sound	can	can opening	–	ours-legacy sound (unverified, model/hand made)
X009	2:49.3–2:52.7	sound	human	food chewing	–	ours-legacy sound (unverified, model/hand made)
X010	2:52.8–2:56.0	sound	human	eating	–	ours-legacy sound (unverified, model/hand made)
X011	2:56.1–2:58.9	sound	human	food chewing	–	ours-legacy sound (unverified, model/hand made)
X012	3:04.7–3:04.9	sound	cutlery	Cutlery impact	–	ours-legacy sound (unverified, model/hand made)
X013	3:06.5–3:07.4	sound	cutlery	Cutlery impact	–	ours-legacy sound (unverified, model/hand made)
X014	3:13.2–3:13.4	sound	cutlery	Cutlery impact	–	ours-legacy sound (unverified, model/hand made)
X015	3:16.2–3:16.9	sound	cutlery	Cutlery impact	–	ours-legacy sound (unverified, model/hand made)
X016	3:21.3–3:28.9	sound	cutlery	Cutlery impact	–	ours-legacy sound (unverified, model/hand made)
X017	3:29.2–3:44.7	sound	utensils	Cutlery sounds	–	ours-legacy sound (unverified, model/hand made)
X018	3:46.8–3:55.6	sound	drink	Drink sound	–	ours-legacy sound (unverified, model/hand made)
X019	3:56.6–3:58.2	sound	utensils	Cutlery sounds	–	ours-legacy sound (unverified, model/hand made)
X020	4:00.3–4:00.6	sound	camera	camera shutter click	–	ours-legacy sound (unverified, model/hand made)

loc2_script3_seq4_rec1

script 3 · window 250 s · speech annos 20(wearer 1 / partner 19)· savvy sound 0 · legacy sound 22 · partner recording: loc2_script3_seq4_rec2

video: SAVVY Video/loc2_script3_seq4_rec1.mp4(20 fps 去畸变小视频,音轨 = mic 5+6 stereo;时间 0 = device 953.25 s)

上条:sub-goal 区段(点击跳转);下条:事件刻度,绿 = wearer 说话,橙 = partner 说话,蓝 = SAVVY sound,灰 = legacy sound

Goal / sub-goal 草稿(Gemini 3.1 Pro,candidate (Gemini draft, not human verified))

activity_type(官方列表选一):dining
overall_goal:Set the table and eat dinner together.
roles:wearer = host, partner = guest
unrelated events:S018, S019 · validation errors: 0

#	sub-goal	clip time	initiated_by	completion	audio_necessity	events	rationale
g1	Gather food, drinks, and tableware for the meal.	0:13.7–1:28.5	partner_request	mixed	speech_only	X000 X001 X002 S000 X003 S001 S002 S003 X004 S004 X005 X006 X007 S005 X008 S006	The partner announces bringing the food, forks, and napkins, and asks the wearer to grab drinks, accompanied by sounds of drawers and cutlery.
g2	Eat the meal and discuss the food quality.	1:42.2–4:07.7	joint_decision	mixed	speech_only	X009 S007 S008 S009 S010 S011 X010 S012 X011 X012 S013 X013 S014 S015 S016 X014 X015 S017 X016 X017 X018 X019 X020 X021	They eat the food, discuss the doneness of the meat, realize they need a knife, and the partner compliments the wearer's cooking.

事件时间线(只用 speech + sound,不用 narration)

id	clip time	kind	speaker / object	text	partner direction · distance	source
X000	0:13.7–0:14.7	sound	Cabinet	Cabinet opening	–	ours-legacy sound (unverified, model/hand made)
X001	0:15.5–0:16.2	sound	Cabinet	Cabinet closing	–	ours-legacy sound (unverified, model/hand made)
X002	0:19.6–0:20.8	sound	Drawer	Drawer opening	–	ours-legacy sound (unverified, model/hand made)
S000	0:23.6–0:24.4	speech	partner	We'll just grab the one... topic: specifying grabbing a particular plate	front-right · 0.57 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X003	0:24.2–0:24.9	sound	Drawer	Drawer closing	–	ours-legacy sound (unverified, model/hand made)
S001	0:27.5–0:28.6	speech	partner	I'll bring the the food. topic: volunteering to bring the food	back · 1.45 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S002	0:30.7–0:31.2	speech	partner	Um couple of forks. topic: mentioning needing forks	back-left · 3.31 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S003	0:37.2–0:38.8	speech	partner	Alright and then if you want to grab a couple of drinks. topic: suggesting grabbing drinks	front · 2.66 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X004	0:58.2–0:58.9	sound	refrigerator	refrigerator door closing	–	ours-legacy sound (unverified, model/hand made)
S004	1:05.6–1:07.1	speech	partner	I'll get you some napkins too. topic: offering to get napkins	front-left · 0.53 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X005	1:06.1–1:06.5	sound	cutlery	Cutlery clinking	–	ours-legacy sound (unverified, model/hand made)
X006	1:12.9–1:13.1	sound	cutlery	Cutlery clinking	–	ours-legacy sound (unverified, model/hand made)
X007	1:14.3–1:23.1	sound	cutlery	Cutlery clinking	–	ours-legacy sound (unverified, model/hand made)
S005	1:21.5–1:22.4	speech	partner	I mean, I trust you. topic: expressing trust in the cook	front-left · 0.56 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X008	1:24.3–1:28.5	sound	cutlery	Cutlery clinking	–	ours-legacy sound (unverified, model/hand made)
S006	1:26.2–1:27.4	speech	partner	I'll give you as much as... topic: stating intention to serve a portion of food	front-left · 0.56 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X009	1:42.2–1:45.5	sound	eating	Eating sounds	–	ours-legacy sound (unverified, model/hand made)
S007	1:46.6–1:47.4	speech	partner	How's the meat. topic: asking about meat quality	front-left · 5.24 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S008	1:48.4–1:49.2	speech	wearer	A little underdone. topic: describing meat as slightly underdone	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S009	1:49.7–1:51.1	speech	partner	Well, that's not terrible. topic: commenting that underdone meat is acceptable	front-left · 5.3 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S010	1:51.2–1:52.3	speech	partner	It's not ground beef. topic: differentiating meat type from ground beef regarding doneness	front-left · 5.09 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S011	1:54.7–1:56.6	speech	partner	I mean, I like beef rare. topic: expressing preference for rare beef	front-left · 5.34 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X010	2:01.0–2:01.4	sound	person	Chewing	–	ours-legacy sound (unverified, model/hand made)
S012	2:08.5–2:09.6	speech	partner	That's always a little tough. topic: commenting on meat toughness	front · 0.79 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X011	2:18.2–2:18.6	sound	person	Chewing	–	ours-legacy sound (unverified, model/hand made)
X012	2:23.6–2:24.0	sound	person	Chewing	–	ours-legacy sound (unverified, model/hand made)
S013	2:40.4–2:41.2	speech	partner	I like the pepper. topic: expressing liking for the pepper	front-right · 0.91 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X013	2:43.7–2:43.9	sound	can	Opening can	–	ours-legacy sound (unverified, model/hand made)
S014	3:05.5–3:06.5	speech	partner	It's got a knife. What am I doing? topic: realizing having a knife while using hands	front-right · 0.88 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S015	3:09.6–3:10.4	speech	partner	Did I give you one? topic: asking if a knife was provided	front · 0.9 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S016	3:22.3–3:25.4	speech	partner	It's actually. Just slightly pink. It's alright. topic: assessing meat doneness (slightly pink) as acceptable	front-right · 0.96 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X014	3:26.6–3:28.7	sound	knife	cut	–	ours-legacy sound (unverified, model/hand made)
X015	3:30.3–3:47.6	sound	cutlery	Cutlery sounds.	–	ours-legacy sound (unverified, model/hand made)
S017	3:53.8–3:54.7	speech	partner	Good cooking, Michael. topic: complimenting Michael's cooking	front-right · 1.02 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S018	3:58.8–3:59.7	speech	partner	Where's my phone? topic: asking about phone location	front-right · 1.03 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X016	3:59.3–4:00.8	sound	Cutlery	Cutlery sounds	–	ours-legacy sound (unverified, model/hand made)
X017	4:01.2–4:02.0	sound	Cutlery	Cutlery sounds	–	ours-legacy sound (unverified, model/hand made)
X018	4:02.5–4:03.0	sound	Cutlery	Cutlery sounds	–	ours-legacy sound (unverified, model/hand made)
X019	4:03.4–4:03.8	sound	Cutlery	Cutlery sounds	–	ours-legacy sound (unverified, model/hand made)
S019	4:04.7–4:06.3	speech	partner	We can stop this. topic: suggesting stopping the recording	front · 2.24 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X020	4:04.8–4:05.7	sound	Cutlery	Cutlery sounds	–	ours-legacy sound (unverified, model/hand made)
X021	4:07.3–4:07.7	sound	Cutlery	Cutlery sounds	–	ours-legacy sound (unverified, model/hand made)

Catching up at home (pair, both views)

loc1_script2_seq7_rec1

script 2 · window 265 s · speech annos 26(wearer 8 / partner 18)· savvy sound 0 · legacy sound 4 · partner recording: loc1_script2_seq7_rec2

video: SAVVY Video/loc1_script2_seq7_rec1.mp4(20 fps 去畸变小视频,音轨 = mic 5+6 stereo;时间 0 = device 2305.40 s)

上条:sub-goal 区段(点击跳转);下条:事件刻度,绿 = wearer 说话,橙 = partner 说话,蓝 = SAVVY sound,灰 = legacy sound

Goal / sub-goal 草稿(Gemini 3.1 Pro,candidate (Gemini draft, not human verified))

activity_type(官方列表选一):multi-person indoor activity (games, conversation)
overall_goal:Select and watch a movie together on the TV.
roles:wearer = host, partner = guest
unrelated events:S000, S001, S002, S003, S004, S005, S006, S007, S008 · validation errors: 0

#	sub-goal	clip time	initiated_by	completion	audio_necessity	events	rationale
g1	Select a movie to watch on the TV.	0:22.0–2:56.7	joint_decision	speech	speech_only	S009 S010 X000 X001 S011 S012 S013 S014 S015	The wearer asks what they should watch, reads a content warning, and they discuss the movie's soundtrack before confirming they chose Shrek.
g2	Adjust the volume of the TV.	2:59.8–3:11.5	wearer_self	mixed	speech_only	X002 S016 S017	The wearer states the TV is too quiet for them, prompting an adjustment of the volume.
g3	Watch the movie while reciting its opening lines.	3:12.8–4:15.9	joint_decision	speech	speech_only	S018 S019 X003 S020 S021 S022 S023 S024 S025	The partner recites the opening narration of Shrek aloud as the movie begins playing.

事件时间线(只用 speech + sound,不用 narration)

id	clip time	kind	speaker / object	text	partner direction · distance	source
S000	0:00.3–0:00.9	speech	wearer	Alright, cool. topic: acknowledging agreement	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S001	0:17.9–0:18.6	speech	partner	I'm very curious. topic: expressing curiosity about the chosen show	back-right · 0.68 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X000	0:22.0–0:22.4	sound	remote	turn on tv	–	ours-legacy sound (unverified, model/hand made)
S002	0:37.3–0:39.0	speech	partner	Yeah, they're trying to upload. topic: confirming upload attempts	back-right · 0.68 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S003	0:43.3–0:44.8	speech	wearer	Is it going fast, did they say? topic: asking about reported upload speed	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S004	0:45.6–0:48.2	speech	partner	It was going. It was definitely faster than the other place. topic: comparing upload speed favorably to another location	front-right · 0.58 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S005	0:48.8–0:51.0	speech	partner	Like recordings were getting uploaded. topic: clarifying previous successful uploads	back-right · 0.69 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S006	0:52.6–0:57.0	speech	wearer	Yeah, the other ones were just like, I recorded the same one or I uploaded the same one all day and it didn't finish. topic: describing a past failed all-day upload attempt	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S007	1:17.9–1:20.2	speech	wearer	No, I don't I don't actually think you're like George. topic: denying listener's resemblance to George	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S008	1:21.0–1:22.6	speech	partner	Really? Okay. No. That's good. topic: expressing surprised relief about not resembling George	back-right · 0.65 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S009	1:29.6–1:31.1	speech	wearer	What can we watch that we both? topic: seeking mutually enjoyable show suggestions	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S010	1:52.1–1:55.5	speech	wearer	One brief drug reference and occasional suggestive content. topic: reciting movie content advisory details	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X001	2:02.6–2:10.3	sound	tv	tv	–	ours-legacy sound (unverified, model/hand made)
S011	2:06.3–2:09.1	speech	partner	We take it to to all star. topic: associating movie music with "All Star" song	back-right · 0.69 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S012	2:13.0–2:16.2	speech	partner	This one's The Counting Crows. We'll take it to the Counting Crows maybe. topic: identifying Shrek opening band as Counting Crows	front-right · 0.53 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S013	2:26.2–2:27.0	speech	partner	I think that's the song. topic: guessing identification of the song	front-right · 0.56 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S014	2:40.3–2:43.2	speech	partner	I don't remember the very beginning. topic: stating inability to recall the movie's start	back-right · 0.7 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S015	2:54.6–2:56.7	speech	wearer	They're going to see us watching Shrek and be like... topic: predicting judgment for watching Shrek	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X002	2:59.8–3:00.4	sound	remote control	sound of TV being adjusted	–	ours-legacy sound (unverified, model/hand made)
S016	3:08.6–3:09.2	speech	partner	I can hear it. topic: confirming audio is audible	front-right · 0.52 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S017	3:10.6–3:11.5	speech	wearer	It's a little quiet for me. topic: stating the audio volume is too low	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S018	3:12.8–3:16.0	speech	partner	They sorted that it was cursed with a frightful enchantment. topic: (TV) narrating the princess's enchantment	front · 2.49 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S019	3:21.3–3:31.9	speech	partner	Desperate, they sought the help of a fairy godmother who had to lock the young princess away in a tower there to awake the kiss of the handsome prince Charles. topic: (TV) narrating fairy godmother's solution and princess's imprisonment awaiting rescue	front · 2.48 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X003	3:29.3–3:59.0	sound	TV	Movie playing on TV	–	ours-legacy sound (unverified, model/hand made)
S020	3:34.3–3:45.2	speech	partner	It was he who would chance a terrorist journey through blistering cold and scorching desert crowning for many days and nights. Risking life and lives. topic: (TV) narrating the prince's arduous journey	front · 2.48 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S021	3:45.2–3:48.7	speech	partner	To reach the dragons. topic: (TV) narrating the goal of reaching dragons	front · 2.48 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S022	3:52.2–3:58.5	speech	partner	For he was the bravest. Most cancer. topic: (TV) narrating the prince's bravery and appearance	front · 2.47 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S023	4:02.6–4:07.4	speech	partner	And it was destiny that his kiss. topic: (TV) narrating the prince's destined kiss action	front · 2.48 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S024	4:07.7–4:10.1	speech	partner	Break the dreaded curse. topic: (TV) narrating the purpose of the kiss breaking the curse	front · 2.48 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S025	4:10.6–4:15.9	speech	partner	Here we climb to the highest room of the tallest town. topic: (TV) narrating the climb to the highest tower room	front · 2.48 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)

loc1_script2_seq7_rec2

script 2 · window 270 s · speech annos 28(wearer 0 / partner 28)· savvy sound 0 · legacy sound 7 · partner recording: loc1_script2_seq7_rec1

video: SAVVY Video/loc1_script2_seq7_rec2.mp4(20 fps 去畸变小视频,音轨 = mic 5+6 stereo;时间 0 = device 3123.20 s)

上条:sub-goal 区段(点击跳转);下条:事件刻度,绿 = wearer 说话,橙 = partner 说话,蓝 = SAVVY sound,灰 = legacy sound

Goal / sub-goal 草稿(Gemini 3.1 Pro,candidate (Gemini draft, not human verified))

activity_type(官方列表选一):multi-person indoor activity (games, conversation)
overall_goal:Watch a movie together and converse
roles:wearer = guest, partner = host
unrelated events:S004, S005, S017, S027 · validation errors: 0

#	sub-goal	clip time	initiated_by	completion	audio_necessity	events	rationale
g1	Decide to watch TV and select a movie.	0:01.1–2:10.7	partner_request	mixed	speech_only	S000 X000 S001 X001 S002 S003 S006 S007 S008 S009 S010 S011 X002 X003 X004 X005	The partner offers to watch TV, discusses what to watch, and uses the remote to select a movie.
g2	Discuss the movie Shrek and adjust the volume.	2:17.9–3:12.2	unknown	speech	speech_only	S012 S013 S014 S015 S016 S018	The partner talks about the movie Shrek, compliments the wearer's trivia knowledge, and notes the volume is low.
g3	Watch the movie and recite the opening narration.	3:13.5–4:16.6	unknown	mixed	speech_only	S019 S020 X006 S021 S022 S023 S024 S025 S026	The partner recites the opening lines of the movie as it begins playing on the TV.

事件时间线(只用 speech + sound,不用 narration)

id	clip time	kind	speaker / object	text	partner direction · distance	source
S000	0:01.1–0:01.6	speech	partner	Alright, cool. topic: acknowledging agreement	front · 1.77 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X000	0:01.4–0:04.5	sound	person	footstep	–	ours-legacy sound (unverified, model/hand made)
S001	0:02.2–0:03.3	speech	partner	Did you want to watch some TV? topic: proposing watching TV	front · 0.94 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X001	0:05.5–0:10.4	sound	person	footstep	–	ours-legacy sound (unverified, model/hand made)
S002	0:12.8–0:14.1	speech	partner	The show made me think of you. topic: connecting the show choice to the listener	front-left · 0.57 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S003	0:16.6–0:17.2	speech	partner	Wait til you see what I chose. topic: hinting at the specific show chosen	front-left · 0.62 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S004	0:44.0–0:45.5	speech	partner	Is it going fast, did they say? topic: asking about reported upload speed	front-left · 0.68 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S005	0:53.4–0:57.8	speech	partner	Yeah, the other ones were just like, I recorded the same one or I uploaded the same one all day and it didn't finish. topic: describing a past failed all-day upload attempt	front-left · 0.66 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S006	1:18.7–1:21.0	speech	partner	No, I don't I don't actually think you're like George. topic: denying listener's resemblance to George	front-left · 0.64 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S007	1:30.3–1:31.8	speech	partner	What can we watch that we both? topic: seeking mutually enjoyable show suggestions	front-left · 0.65 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S008	1:34.8–1:35.9	speech	partner	This movie's actually dope. topic: endorsing the movie as excellent	front-left · 0.62 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S009	1:39.6–1:48.1	speech	partner	I watched this movie recently and we were all like, oh my god, this movie's so good. Like the music is good. It's funnier than I remember. topic: praising movie's music and humor based on recent viewing	front-left · 0.59 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S010	1:39.6–1:48.1	speech	partner	I watched this movie recently and we are all like, oh my god, this movie's so good. Like the music is good. It's funnier than I remember. topic: praising movie's music and humor based on recent viewing	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S011	1:52.9–1:56.3	speech	partner	One brief drug reference and occasional suggestive content. topic: reciting movie content advisory details	front-left · 0.7 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X002	1:59.2–1:59.3	sound	remote control	click	–	ours-legacy sound (unverified, model/hand made)
X003	2:03.0–2:03.1	sound	remote control	click	–	ours-legacy sound (unverified, model/hand made)
X004	2:03.3–2:03.4	sound	remote control	click	–	ours-legacy sound (unverified, model/hand made)
X005	2:10.6–2:10.7	sound	remote control	click	–	ours-legacy sound (unverified, model/hand made)
S012	2:17.9–2:21.0	speech	partner	You would know the band that of the opening scene of Shrek. topic: acknowledging listener's specific band knowledge (Shrek opening)	front-left · 0.49 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S013	2:25.4–2:26.4	speech	partner	I would never know that. topic: admitting lack of specific band knowledge	front-left · 0.51 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S014	2:28.6–2:31.3	speech	partner	You do realize you're like an encyclopedia. topic: complimenting listener's encyclopedic knowledge	front-left · 0.55 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S015	2:33.8–2:36.6	speech	partner	Actually a shocking amount of information is in your brain. topic: remarking on listener's surprising amount of knowledge	front-left · 0.58 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S016	2:46.8–2:48.2	speech	partner	Do you like Shrek one or Shrek two better? topic: asking for preference between Shrek sequels	front-left · 0.56 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S017	2:55.3–2:57.4	speech	partner	They're going to see us watching Shrek and be like... topic: predicting judgment for watching Shrek	front-left · 0.67 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S018	3:11.3–3:12.2	speech	partner	It's a little quiet for me. topic: stating the audio volume is too low	front-left · 0.7 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S019	3:13.5–3:16.7	speech	partner	They sorted that it was cursed with a frightful enchantment. topic: (TV) narrating the princess's enchantment	front · 2.97 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S020	3:22.0–3:32.7	speech	partner	Desperate, they sought the help of a fairy godmother who had to lock the young princess away in a tower there to awake the kiss of the handsome prince Charles. topic: (TV) narrating fairy godmother's solution and princess's imprisonment awaiting rescue	front · 2.94 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X006	3:30.1–3:59.7	sound	TV	Movie playing	–	ours-legacy sound (unverified, model/hand made)
S021	3:35.0–3:45.9	speech	partner	It was he who would chance a terrorist journey through blistering cold and scorching desert crowning for many days and nights. Risking life and lives. topic: (TV) narrating the prince's arduous journey	front · 2.94 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S022	3:45.9–3:49.4	speech	partner	To reach the dragons. topic: (TV) narrating the goal of reaching dragons	front · 2.94 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S023	3:52.9–3:59.2	speech	partner	For he was the bravest. Most cancer. topic: (TV) narrating the prince's bravery and appearance	front · 2.94 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S024	4:03.3–4:08.1	speech	partner	And it was destiny that his kiss. topic: (TV) narrating the prince's destined kiss action	front · 2.94 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S025	4:08.4–4:10.8	speech	partner	Break the dreaded curse. topic: (TV) narrating the purpose of the kiss breaking the curse	front · 2.94 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S026	4:11.3–4:16.6	speech	partner	Here we climb to the highest room of the tallest town. topic: (TV) narrating the climb to the highest tower room	front · 2.94 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S027	4:21.2–4:24.0	speech	partner	I think we're good. It's just three to five minutes and we're at 4 minutes. topic: suggesting stopping recording as duration requirement met	front-left · 0.6 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)

Snacks and drinks for a guest

loc2_script2_seq4_rec2

script 2 · window 99 s · speech annos 11(wearer 2 / partner 9)· savvy sound 0 · legacy sound 8 · partner recording: loc2_script2_seq4_rec1

video: SAVVY Video/loc2_script2_seq4_rec2.mp4(20 fps 去畸变小视频,音轨 = mic 5+6 stereo;时间 0 = device 2787.20 s)

上条:sub-goal 区段(点击跳转);下条:事件刻度,绿 = wearer 说话,橙 = partner 说话,蓝 = SAVVY sound,灰 = legacy sound

Goal / sub-goal 草稿(Gemini 3.1 Pro,candidate (Gemini draft, not human verified))

activity_type(官方列表选一):preparing snacks
overall_goal:Prepare popcorn and coffee to enjoy with the guest.
roles:wearer = host, partner = guest
unrelated events:S000 · validation errors: 0

#	sub-goal	clip time	initiated_by	completion	audio_necessity	events	rationale
g1	Fetch creamer for the coffee.	0:07.1–0:28.0	partner_request	mixed	speech_only	S001 S002 X000 X001	The partner requests creamer, the wearer agrees, and refrigerator sounds indicate it is being fetched.
g2	Prepare and evaluate the popcorn.	0:16.9–0:46.6	partner_request	mixed	speech_only	S003 S004 X002 S005 S006	The partner suggests popcorn, the wearer confirms, microwave sounds are heard, and they discuss the plant-based choice.
g3	Mix the creamer into the coffee and taste it.	0:55.4–1:37.3	joint_decision	speech	speech_only	S007 X003 X004 X005 S008 S009 X006 X007 S010	The partner mentions shaking the creamer, suggests putting it back, and then they taste the coffee.

事件时间线(只用 speech + sound,不用 narration)

id	clip time	kind	speaker / object	text	partner direction · distance	source
S000	0:00.3–0:01.0	speech	partner	What's up? topic: greeting and asking about status	front-right · 1.2 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S001	0:07.1–0:09.7	speech	partner	So I think we need some a creamer. topic: stating need for creamer	front-right · 1.27 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S002	0:13.7–0:14.8	speech	wearer	I'll get you a little creamer. topic: reiterating intention to get creamer	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S003	0:16.9–0:18.4	speech	partner	Yeah. Popcorn will be nice. topic: accepting snack offer (popcorn)	front-right · 2.97 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S004	0:18.7–0:19.9	speech	wearer	Cool. We got a little popcorn. topic: confirming popcorn availability	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X000	0:21.1–0:21.9	sound	refrigerator	refrigerator door opening	–	ours-legacy sound (unverified, model/hand made)
X001	0:27.6–0:28.0	sound	refrigerator	refrigerator door closing	–	ours-legacy sound (unverified, model/hand made)
X002	0:29.5–0:29.6	sound	kitchen appliance	microwave oven	–	ours-legacy sound (unverified, model/hand made)
S005	0:37.8–0:39.0	speech	partner	How's the popcorn? topic: asking about popcorn quality	front-left · 1.93 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S006	0:45.0–0:46.6	speech	partner	Plant based. Interesting choice. topic: commenting on popcorn being plant-based	front-right · 1.36 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S007	0:55.4–0:57.4	speech	partner	I don't think I shook. I don't think I shook this well enough but. topic: expressing doubt about properly shaking creamer	back-left · 4.34 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X003	0:56.3–0:56.6	sound	drawer	drawer open or close	–	ours-legacy sound (unverified, model/hand made)
X004	0:56.6–0:58.8	sound	drawer	drawer open or close	–	ours-legacy sound (unverified, model/hand made)
X005	0:58.8–0:59.2	sound	drawer	drawer open or close	–	ours-legacy sound (unverified, model/hand made)
S008	1:16.8–1:17.5	speech	partner	You want to put that back? topic: asking about putting item back	front · 1.25 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S009	1:24.1–1:25.2	speech	partner	Let's try to coffee now. topic: suggesting trying the coffee	back-right · 3.95 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X006	1:25.2–1:25.8	sound	cupboard	cupboard door opens or closes	–	ours-legacy sound (unverified, model/hand made)
X007	1:27.9–1:28.3	sound	cupboard	cupboard door opens or closes	–	ours-legacy sound (unverified, model/hand made)
S010	1:36.4–1:37.3	speech	partner	Taste pretty good. topic: commenting positively on coffee taste	front-right · 1.03 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)

loc2_script2_seq5_rec1

script 2 · window 255 s · speech annos 24(wearer 0 / partner 24)· savvy sound 0 · legacy sound 14 · partner recording: loc2_script2_seq5_rec2

video: SAVVY Video/loc2_script2_seq5_rec1.mp4(20 fps 去畸变小视频,音轨 = mic 5+6 stereo;时间 0 = device 1297.21 s)

上条:sub-goal 区段(点击跳转);下条:事件刻度,绿 = wearer 说话,橙 = partner 说话,蓝 = SAVVY sound,灰 = legacy sound

Goal / sub-goal 草稿(Gemini 3.1 Pro,candidate (Gemini draft, not human verified))

activity_type(官方列表选一):multi-person indoor activity (games, conversation)
overall_goal:Play a game of Scrabble and eat snacks with the partner.
roles:wearer = unknown, partner = unknown
unrelated events:S006, S007, S018, S021, S022 · validation errors: 0

#	sub-goal	clip time	initiated_by	completion	audio_necessity	events	rationale
g1	Set up the board game and distribute tiles.	0:02.3–1:07.3	partner_request	speech	speech_only	S000 X000 S001 S002 S003 X001 S004 S005 X002 S008 X003 S009 X004	The partner initiates the game setup, suggesting to open the game and distribute seven letters each.
g2	Decide who goes first and play the first word.	1:08.2–3:11.7	joint_decision	mixed	speech_only	S010 S011 X005 S012 X006 S013 S014 X007 S015 S016 S017 X008 X009	They discuss who will go first, and the partner plays the word 'Juke', then draws replacement tiles.
g3	Continue playing words and conclude the game.	3:36.1–4:15.1	joint_decision	mixed	speech_only	X010 S019 X011 X012 S020 X013 S023	The wearer plays a turn, the partner plays the word 'Gore', and they conclude the game while eating snacks.

事件时间线(只用 speech + sound,不用 narration)

id	clip time	kind	speaker / object	text	partner direction · distance	source
S000	0:02.3–0:03.1	speech	partner	Ready to start? topic: asking about readiness to start	front-right · 1.52 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X000	0:10.4–0:16.5	sound	Bag	Unpacking a brown paper bag.	–	ours-legacy sound (unverified, model/hand made)
S001	0:23.9–0:25.7	speech	partner	Alright, let's put this aside. topic: suggesting putting item aside	front-right · 0.89 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S002	0:29.6–0:30.7	speech	partner	Open the game up. topic: instructing opening the game	front-right · 0.84 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S003	0:32.4–0:33.6	speech	partner	Give you one of these. topic: giving a game piece holder	front-right · 0.8 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X001	0:33.6–0:34.4	sound	popcorn	popcorn	–	ours-legacy sound (unverified, model/hand made)
S004	0:34.9–0:36.3	speech	partner	I'll take one of these. topic: taking a game piece holder	front-right · 0.78 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S005	0:37.4–0:40.1	speech	partner	Give me some letters, I'll give you 7 letters, I believe. topic: stating plan to distribute seven letters	front-right · 0.83 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S006	0:47.0–0:48.1	speech	partner	We're recording. topic: announcing recording is active	front-right · 1.0 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S007	0:50.7–0:52.2	speech	partner	Uh few more minutes. topic: indicating need for a few more minutes	front-right · 0.93 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X002	0:54.5–0:55.1	sound	scrabble tiles	scrabble	–	ours-legacy sound (unverified, model/hand made)
S008	1:00.2–1:01.0	speech	partner	Yeah, let's do it. topic: agreeing to place words	front-right · 0.94 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X003	1:01.2–1:03.8	sound	popcorn	Popcorn popping	–	ours-legacy sound (unverified, model/hand made)
S009	1:02.8–1:04.2	speech	partner	Alright, let's get set up. topic: proposing getting set up	front-right · 0.95 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X004	1:05.6–1:07.3	sound	popcorn	Popcorn popping	–	ours-legacy sound (unverified, model/hand made)
S010	1:08.2–1:09.2	speech	partner	I can go first. topic: offering to take first turn	front-right · 0.85 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S011	1:20.6–1:23.0	speech	partner	Alright, let's see what words do I have here. topic: considering words from available letters	front-right · 0.92 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X005	1:29.2–1:29.2	sound	game	Scrabble	–	ours-legacy sound (unverified, model/hand made)
S012	1:38.9–1:39.9	speech	partner	You want to go first? topic: offering first turn to listener	front-right · 0.93 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X006	1:42.2–2:00.0	sound	paper bag	crumpling	–	ours-legacy sound (unverified, model/hand made)
S013	1:46.6–1:47.6	speech	partner	I actually go first. topic: stating taking the first turn	front-right · 0.91 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S014	1:53.8–1:55.4	speech	partner	Here we go, we got Juke. topic: announcing playing the word "Juke"	front-right · 0.8 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X007	2:00.7–2:04.3	sound	Eating sounds	Eating	–	ours-legacy sound (unverified, model/hand made)
S015	2:06.2–2:08.2	speech	partner	Don't worry about it, man, I mean, it means something. topic: asserting word validity without defining	front-right · 0.68 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S016	2:11.0–2:11.5	speech	partner	It's appropriate. topic: confirming word appropriateness	front-right · 0.64 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S017	2:22.7–2:25.5	speech	partner	Hmm, should take some tiles for once I just took. topic: realizing need for drawing replacement tiles	front-right · 0.83 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X008	2:55.4–2:56.0	sound	paper bag	Rustling	–	ours-legacy sound (unverified, model/hand made)
X009	3:02.6–3:11.7	sound	game	scrabble	–	ours-legacy sound (unverified, model/hand made)
S018	3:12.0–3:14.1	speech	partner	Alright, you think we're good or if we keep playing a little bit? topic: inquiring whether to stop or continue playing	front-right · 0.76 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X010	3:36.1–3:37.3	sound	Hand	Placing game piece	–	ours-legacy sound (unverified, model/hand made)
S019	3:38.8–3:39.9	speech	partner	There you go, Ron. topic: acknowledging Ron's play	front-right · 0.83 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X011	3:45.8–3:46.7	sound	Eating popcorn	Crunch	–	ours-legacy sound (unverified, model/hand made)
X012	3:54.0–3:54.9	sound	Eating popcorn	Crunch	–	ours-legacy sound (unverified, model/hand made)
S020	3:55.3–3:56.0	speech	partner	Gore. topic: playing the word "Gore"	front-right · 0.85 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X013	4:06.6–4:11.3	sound	Tile	Scratching	–	ours-legacy sound (unverified, model/hand made)
S021	4:08.0–4:08.6	speech	partner	Not yet. topic: indicating something is not ready	front-right · 0.89 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S022	4:09.6–4:12.4	speech	partner	There it did, we should work probably about good here. topic: suggesting concluding the recording/play	front-right · 0.89 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S023	4:13.8–4:15.1	speech	partner	Cool, good game. topic: praising the game	front-right · 0.84 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)

loc2_script2_seq2_rec1

script 2 · window 106 s · speech annos 18(wearer 4 / partner 14)· savvy sound 0 · legacy sound 5 · partner recording: loc2_script2_seq2_rec2

video: SAVVY Video/loc2_script2_seq2_rec1.mp4(20 fps 去畸变小视频,音轨 = mic 5+6 stereo;时间 0 = device 398.20 s)

上条:sub-goal 区段(点击跳转);下条:事件刻度,绿 = wearer 说话,橙 = partner 说话,蓝 = SAVVY sound,灰 = legacy sound

Goal / sub-goal 草稿(Gemini 3.1 Pro,candidate (Gemini draft, not human verified))

activity_type(官方列表选一):touring the room
overall_goal:Take a guided tour of the host's home
roles:wearer = guest, partner = host
unrelated events:none · validation errors: 0

#	sub-goal	clip time	initiated_by	completion	audio_necessity	events	rationale
g1	Initiate the tour and view the main living areas	0:05.7–0:38.7	joint_decision	speech	speech_only	S000 X000 S001 S002 S003 S004 S005	The partner suggests starting the tour in S000 and points out a picture, the kitchen, and the pool in S001 to S004, which the wearer acknowledges in S005.
g2	Discuss the kitchen activities and items	0:40.0–0:57.9	wearer_self	speech	speech_only	S006 S007 S008 X001 S009	The wearer notices signs of cooking in S006, prompting the partner to explain their recent meal and show items like rum in S007 to S009.
g3	Continue the tour through the remaining rooms and features	0:58.9–1:31.2	wearer_self	speech	speech_only	S010 S011 X002 S012 X003 X004 S013 S014 S015	The wearer compliments the paintings in S010, and the partner proceeds to point out the bathroom, hallway, and guest room in S011 to S015.
g4	Conclude the tour	1:37.2–1:43.7	partner_request	speech	speech_only	S016 S017	The partner suggests heading back in S016, and the wearer agrees in S017, ending the tour.

事件时间线(只用 speech + sound,不用 narration)

id	clip time	kind	speaker / object	text	partner direction · distance	source
S000	0:05.7–0:06.9	speech	partner	Yeah yeah. Let's do the tour. topic: agreeing to conduct the requested tour	front-left · 1.2 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X000	0:12.1–0:12.4	sound	human	footsteps	–	ours-legacy sound (unverified, model/hand made)
S001	0:18.4–0:21.2	speech	partner	Alright and we got a picture of a donkey here which is pretty dope. topic: describing a donkey picture during the tour	front-left · 0.69 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S002	0:21.6–0:22.9	speech	partner	Took that one myself. topic: stating having personally taken the donkey photo	front · 0.77 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S003	0:24.0–0:25.2	speech	partner	Got the kitchen. topic: identifying the kitchen area on the tour	front · 0.79 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S004	0:32.7–0:36.4	speech	partner	The pool. That's the Peace Day Resistance. topic: highlighting the pool as the centerpiece (pièce de résistance)	back-right · 1.18 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S005	0:38.0–0:38.7	speech	wearer	Looks good. topic: commenting positively on the pool's appearance	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S006	0:40.0–0:41.7	speech	wearer	Seem like you're just cooking here recently. topic: observing evidence of recent cooking activity	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S007	0:42.0–0:43.5	speech	partner	Yeah I just made myself a little food. topic: confirming having recently cooked food for oneself	back-left · 0.91 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S008	0:51.0–0:52.2	speech	partner	We got this in there. topic: indicating an item ('this') in the kitchen area	front-left · 0.99 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X001	0:51.5–0:51.8	sound	door	Door opening or closing	–	ours-legacy sound (unverified, model/hand made)
S009	0:56.6–0:57.9	speech	partner	Got the rum right there. topic: identifying the location of the rum	back-left · 1.39 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S010	0:58.9–0:59.9	speech	wearer	I like the paintings. topic: expressing appreciation for the paintings	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S011	1:02.4–1:05.0	speech	partner	That's from the Amazon. topic: specifying painting origin as the Amazon	back-right · 1.14 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X002	1:05.1–1:05.4	sound		footstep	–	ours-legacy sound (unverified, model/hand made)
S012	1:10.7–1:12.2	speech	partner	Yeah we got that room there. topic: identifying another room ('that room') during the tour	back-left · 1.22 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X003	1:16.1–1:16.3	sound		footstep	–	ours-legacy sound (unverified, model/hand made)
X004	1:17.0–1:17.1	sound		object tap	–	ours-legacy sound (unverified, model/hand made)
S013	1:22.6–1:23.2	speech	partner	Bathroom. topic: naming the bathroom location	back · 1.06 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S014	1:23.5–1:24.3	speech	partner	Hallway. topic: naming the hallway location	back · 1.08 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S015	1:30.2–1:31.2	speech	partner	Guest room over there. topic: pointing out the location of the guest room	back-right · 0.76 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S016	1:37.2–1:38.1	speech	partner	Let's take it back here. topic: suggesting returning to the initial area	back · 0.98 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S017	1:42.9–1:43.7	speech	wearer	Oh, we're good. topic: indicating readiness to end the interaction	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)

loc2_script2_seq1_rec1

script 2 · window 134 s · speech annos 15(wearer 3 / partner 12)· savvy sound 1 · legacy sound 9 · partner recording: loc2_script2_seq1_rec2

video: SAVVY Video/loc2_script2_seq1_rec1.mp4(20 fps 去畸变小视频,音轨 = mic 5+6 stereo;时间 0 = device 121.32 s)

上条:sub-goal 区段(点击跳转);下条:事件刻度,绿 = wearer 说话,橙 = partner 说话,蓝 = SAVVY sound,灰 = legacy sound

Goal / sub-goal 草稿(Gemini 3.1 Pro,candidate (Gemini draft, not human verified))

activity_type(官方列表选一):multi-person indoor activity (games, conversation)
overall_goal:Enter the host's place, get a drink, and catch up.
roles:wearer = guest, partner = host
unrelated events:S007, S008, S011 · validation errors: 0

#	sub-goal	clip time	initiated_by	completion	audio_necessity	events	rationale
g1	Enter the host's home and settle in.	0:02.0–0:14.2	wearer_self	speech	audio_helps	X000 A000 X001 X002 S000 S001	The wearer knocks on the door and enters, prompting the partner to invite them to take a seat in S000.
g2	Get a drink from the host.	0:20.6–2:07.8	partner_request	sound	audio_helps	S002 S003 X003 S004 X004 S006 X006 X007 X008	The partner offers a drink in S002, the wearer asks them to hurry in S004, and sounds of cans opening and drinking confirm the action.
g3	Catch up and converse.	0:49.0–2:08.2	wearer_self	speech	speech_only	S005 X005 S009 S010 S012 S013 S014	The wearer initiates small talk about a goal in S005, leading to a conversation about how they have been, the AC, and the weather.

事件时间线(只用 speech + sound,不用 narration)

id	clip time	kind	speaker / object	text	partner direction · distance	source
X000	0:02.0–0:04.3	sound	door	knocking on door	–	ours-legacy sound (unverified, model/hand made)
A000	0:02.4–0:04.4	sound	door	knocking door	front-right · 0.52 m	savvy-bench sound (QA-selected event)
X001	0:04.8–0:06.9	sound	door	door opening	–	ours-legacy sound (unverified, model/hand made)
X002	0:10.1–0:11.3	sound	door	door closing	–	ours-legacy sound (unverified, model/hand made)
S000	0:11.4–0:12.1	speech	partner	Yeah, take a seat. topic: inviting to sit down	back-left · 3.16 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S001	0:13.4–0:14.2	speech	wearer	I like your door. topic: expressing liking for the door	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S002	0:20.6–0:22.1	speech	partner	Hey, you want anything to drink? topic: asking about drink preference	front-left · 0.8 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S003	0:38.7–0:39.4	speech	partner	Yeah, I don't know. topic: reaffirming uncertainty about flavor	front · 1.76 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X003	0:40.6–0:45.7	sound		footsteps	–	ours-legacy sound (unverified, model/hand made)
S004	0:44.6–0:45.4	speech	wearer	Hurry please. topic: requesting faster service	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X004	0:46.1–0:46.3	sound		door closing	–	ours-legacy sound (unverified, model/hand made)
S005	0:49.0–0:50.3	speech	wearer	That's a nice goal you got. topic: complimenting a goal item	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X005	1:12.8–1:13.0	sound		thunk	–	ours-legacy sound (unverified, model/hand made)
S006	1:15.9–1:17.3	speech	partner	Hey, you want to take a drink of it? topic: prompting to take a drink	front-left · 1.02 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S007	1:21.0–1:22.5	speech	partner	Your thing's working. I think so. topic: confirming belief that device is working	front-left · 1.0 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S008	1:26.2–1:27.5	speech	partner	I think this is it. I think we ended here. topic: proposing the conversation is ending	front-left · 1.08 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S009	1:31.8–1:32.6	speech	partner	Yes, how you been? topic: inquiring about listener's status	front-left · 1.09 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S010	1:36.7–1:37.3	speech	partner	Yeah, same here. topic: expressing shared feeling about day ending	front-left · 1.07 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S011	1:37.4–1:40.7	speech	partner	We just gotta knock out these and we should be all good. topic: expressing need to finish tasks first	front-left · 1.09 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S012	1:48.6–1:50.5	speech	partner	Trynna stay inside, get some AC, man. topic: stating preference for staying indoors with air conditioning	front-left · 1.11 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X006	1:56.8–1:57.5	sound	drink	drinking, gulping	–	ours-legacy sound (unverified, model/hand made)
X007	2:00.8–2:01.7	sound	LaCroix can	Can opening	–	ours-legacy sound (unverified, model/hand made)
S013	2:02.8–2:04.1	speech	partner	Y'all have any weather, yeah. topic: confirming the presence of weather	front-left · 1.12 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S014	2:07.0–2:08.2	speech	partner	Got more weather available. topic: sarcastically commenting on weather availability	front-left · 1.13 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X008	2:07.4–2:07.8	sound	LaCroix can	Can opening	–	ours-legacy sound (unverified, model/hand made)

loc1_script2_seq3_rec1

script 2 · window 196 s · speech annos 14(wearer 1 / partner 13)· savvy sound 1 · legacy sound 13 · partner recording: loc1_script2_seq3_rec2

video: SAVVY Video/loc1_script2_seq3_rec1.mp4(20 fps 去畸变小视频,音轨 = mic 5+6 stereo;时间 0 = device 911.20 s)

上条:sub-goal 区段(点击跳转);下条:事件刻度,绿 = wearer 说话,橙 = partner 说话,蓝 = SAVVY sound,灰 = legacy sound

Goal / sub-goal 草稿(Gemini 3.1 Pro,candidate (Gemini draft, not human verified))

activity_type(官方列表选一):making coffee
overall_goal:Make and serve coffee for the guest
roles:wearer = guest, partner = host
unrelated events:S011, S012 · validation errors: 0

#	sub-goal	clip time	initiated_by	completion	audio_necessity	events	rationale
g1	Offer coffee and locate the coffee cups and supplies.	0:02.5–0:26.3	partner_request	speech	speech_only	S000 S001 S002	The partner offers coffee to the wearer after their flight, and they successfully locate the necessary supplies.
g2	Fill the coffee maker with water and set the brewing amount.	0:44.2–1:21.8	joint_decision	mixed	audio_helps	S003 X000 X001 X002 S004 X003 X004 X005 S005 X006 S006	They discuss adding water, look for a filter, and decide on the amount of water to use while handling the machine and cabinets.
g3	Wait for the coffee to brew and discuss the coffee maker's design.	1:25.3–2:46.6	unknown	mixed	speech_only	S007 S008 S009 S010 X007 A000	While the coffee maker heats up and brews, the partner comments on its sleek and practical design.
g4	Drink the prepared coffee.	2:46.5–3:12.9	unknown	mixed	audio_helps	X008 X009 X010 X011 S013 X012	Sounds of clinking and drinking indicate the coffee is being consumed, followed by a comment about a long sip.

事件时间线(只用 speech + sound,不用 narration)

id	clip time	kind	speaker / object	text	partner direction · distance	source
S000	0:02.5–0:04.2	speech	partner	Would you like some coffee? I know you had a long flight. topic: offering coffee due to long flight	front · 0.78 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S001	0:22.4–0:23.0	speech	wearer	You'll find them. topic: expressing confidence in locating mugs	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S002	0:24.8–0:26.3	speech	partner	Maybe they're right by the coffee. And they are. topic: speculating and confirming mugs are by coffee	front-left · 2.95 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S003	0:44.2–0:45.4	speech	partner	Think we have to add some water. topic: stating belief water needs adding	front-left · 1.09 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X000	0:46.7–0:47.6	sound	cabinet	cabinet open or close	–	ours-legacy sound (unverified, model/hand made)
X001	0:51.4–0:52.2	sound	faucet	water	–	ours-legacy sound (unverified, model/hand made)
X002	0:52.9–0:56.2	sound	faucet	water	–	ours-legacy sound (unverified, model/hand made)
S004	0:57.2–0:58.1	speech	partner	Does she have a water filter? topic: inquiring about availability of a water filter	front-right · 2.29 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X003	1:05.8–1:06.3	sound	cabinet	cabinet open	–	ours-legacy sound (unverified, model/hand made)
X004	1:11.0–1:11.5	sound	cabinet	cabinet close	–	ours-legacy sound (unverified, model/hand made)
X005	1:12.3–1:12.6	sound	cabinet	cabinet open	–	ours-legacy sound (unverified, model/hand made)
S005	1:13.6–1:15.5	speech	partner	Do ten, eight, ten ounces. topic: instructing water amount (8-10 ounces)	front · 1.25 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X006	1:13.7–1:14.0	sound	cabinet	cabinet close	–	ours-legacy sound (unverified, model/hand made)
S006	1:19.9–1:21.8	speech	partner	Here it is. The gear rig. topic: announcing finding the coffee maker	front · 1.7 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S007	1:25.3–1:26.5	speech	partner	I think it just has to heat up. topic: stating belief coffee maker needs heating	front-left · 1.32 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S008	1:32.4–1:33.5	speech	partner	It's gotten so sleek. topic: commenting on the coffee maker's sleekness	front · 1.29 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S009	1:37.1–1:41.6	speech	partner	This design is based on the like big bulky ones just not being super practical. topic: explaining design motivation avoiding bulky models	front-left · 1.12 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S010	1:44.8–1:45.8	speech	partner	Can hide it somewhere. topic: noting the coffee maker can be hidden	front-left · 1.2 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X007	1:55.7–1:58.8	sound	Sink	Water flowing	–	ours-legacy sound (unverified, model/hand made)
A000	1:56.5–2:46.6	sound	coffee maker	water dripping	front · 1.04 m	savvy-bench sound (QA-selected event)
S011	2:21.0–2:22.3	speech	partner	Am I supposed to have a cup too? topic: asking if they also require a cup	front · 1.42 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S012	2:25.6–2:26.9	speech	partner	I don't think we need to do that though. topic: expressing doubt about needing two cups	front · 1.34 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X008	2:46.5–2:47.0	sound	faucet	water sounds	–	ours-legacy sound (unverified, model/hand made)
X009	2:51.5–2:52.0	sound	faucet	water running	–	ours-legacy sound (unverified, model/hand made)
X010	3:02.6–3:03.5	sound	cups	clink	–	ours-legacy sound (unverified, model/hand made)
X011	3:05.7–3:10.5	sound	person	drinking	–	ours-legacy sound (unverified, model/hand made)
S013	3:10.7–3:11.5	speech	partner	Been a long sip. topic: commenting on taking a long sip	front-left · 0.86 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X012	3:12.6–3:12.9	sound	cup	cup placing	–	ours-legacy sound (unverified, model/hand made)

Games and conversation

loc3_script2_seq7_rec1

script 2 · window 100 s · speech annos 25(wearer 12 / partner 13)· savvy sound 0 · legacy sound 1 · partner recording: loc3_script2_seq7_rec2

video: SAVVY Video/loc3_script2_seq7_rec1.mp4(20 fps 去畸变小视频,音轨 = mic 5+6 stereo;时间 0 = device 1144.20 s)

上条:sub-goal 区段(点击跳转);下条:事件刻度,绿 = wearer 说话,橙 = partner 说话,蓝 = SAVVY sound,灰 = legacy sound

Goal / sub-goal 草稿(Gemini 3.1 Pro,candidate (Gemini draft, not human verified))

activity_type(官方列表选一):multi-person indoor activity (games, conversation)
overall_goal:Watch a stock market video and discuss investments with the partner.
roles:wearer = unknown, partner = unknown
unrelated events:S011, S012, S013, S014 · validation errors: 0

#	sub-goal	clip time	initiated_by	completion	audio_necessity	events	rationale
g1	Settle at the couch and introduce the stock market video.	0:05.6–0:42.8	partner_request	speech	speech_only	S000 S001 X000 S002 S003 S004 S005 S006 S007 S008	The partner asks what they are doing at the couch, and the wearer introduces a stock market video, joking about its length.
g2	Watch the video and discuss personal investments.	0:43.5–1:33.3	partner_request	speech	speech_only	S009 S010 S015 S016 S017 S018 S019 S020 S021 S022 S023 S024	While the video plays in the background, the partner asks about the wearer's investments, leading to a discussion about specific companies.

事件时间线(只用 speech + sound,不用 narration)

id	clip time	kind	speaker / object	text	partner direction · distance	source
S000	0:05.6–0:06.8	speech	partner	What're we gonna do at the couch? topic: asking about couch activity	front-right · 1.44 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S001	0:13.0–0:15.1	speech	partner	That thing of that video of that thing. topic: referencing a specific video vaguely	front-right · 0.5 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X000	0:14.4–0:28.5	sound	human	Conversation	–	ours-legacy sound (unverified, model/hand made)
S002	0:15.1–0:17.8	speech	wearer	We got this video about like the stock market. topic: specifying video topic as stock market	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S003	0:23.8–0:25.7	speech	partner	See? This is how to become a tycoon. topic: commenting video teaches becoming tycoon	front-right · 0.31 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S004	0:25.7–0:27.6	speech	wearer	Exactly. You gotta wear that pink shirt. topic: agreeing and joking about required attire	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S005	0:33.9–0:36.8	speech	wearer	It's a very short video. It's only 57 minutes long. topic: sarcastically describing video length	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S006	0:37.5–0:39.9	speech	wearer	University takes takes four years to get a bachelor's degree. topic: comparing video length to university degree duration	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S007	0:40.1–0:41.4	speech	wearer	This is only an hour so. topic: stating video is about an hour long	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S008	0:42.0–0:42.8	speech	wearer	It's pretty quick. topic: calling video duration quick comparatively	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S009	0:43.5–0:47.6	speech	partner	If you're not a fan of this kind of setup or if you like the whiteboard better. topic: acknowledging viewer preference for video setup	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S010	0:47.6–0:50.1	speech	partner	Um the reason I'm doing it this way is because this is what. topic: explaining reason for video presentation style	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S011	0:50.1–0:51.9	speech	wearer	Do you have a video you want to show me too? topic: asking if other person has video to show	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S012	0:52.7–0:54.7	speech	wearer	Is that what we're supposed to do? We both show each other videos. topic: questioning if task involves mutual video sharing	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S013	0:54.9–0:57.9	speech	partner	I think you're supposed to show me a video that reminds you of me. topic: stating belief task is showing reminding video	front-right · 0.36 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S014	0:59.4–1:01.8	speech	partner	That's what it says in the script but it doesn't actually have to. topic: mentioning script instruction about video choice	front-right · 0.36 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S015	1:01.8–1:04.1	speech	partner	There we go. Just watch this for like 15 more seconds topic: suggesting watching current video longer	front-right · 0.34 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S016	1:04.1–1:06.7	speech	partner	Let's get right into it. So, this presentation here. topic: introducing video presentation content	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S017	1:06.7–1:08.1	speech	partner	Are you invested in anything? topic: asking about personal investments	front-right · 0.36 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S018	1:11.2–1:13.3	speech	wearer	I bought some Endeavor. topic: stating purchase of Endeavor stock	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S019	1:13.3–1:17.3	speech	wearer	That's like the company that owns like the UFC and stuff like that. topic: explaining Endeavor owns UFC	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S020	1:18.1–1:22.8	speech	partner	I'm 33 at the time of this recording which is pretty crazy how time flies. topic: stating age and commenting on time passing	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S021	1:22.8–1:25.3	speech	partner	Um but I started with a trade king account. topic: mentioning starting with TradeKing account	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S022	1:25.3–1:28.4	speech	wearer	You're not loved by the public but like they're very profitable. topic: commenting on Endeavor's profitability despite perception	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S023	1:28.4–1:30.3	speech	wearer	So, I think they'll go up. topic: predicting Endeavor stock will increase	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S024	1:30.7–1:33.3	speech	partner	Trade and one finance Robin Hood. topic: mentioning other trading platforms	–	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)

loc3_script3_seq2_rec1

script 3 · window 169 s · speech annos 22(wearer 0 / partner 22)· savvy sound 0 · legacy sound 3 · partner recording: loc3_script3_seq2_rec2

video: SAVVY Video/loc3_script3_seq2_rec1.mp4(20 fps 去畸变小视频,音轨 = mic 5+6 stereo;时间 0 = device 1749.22 s)

上条:sub-goal 区段(点击跳转);下条:事件刻度,绿 = wearer 说话,橙 = partner 说话,蓝 = SAVVY sound,灰 = legacy sound

Goal / sub-goal 草稿(Gemini 3.1 Pro,candidate (Gemini draft, not human verified))

activity_type(官方列表选一):multi-person indoor activity (games, conversation)
overall_goal:Sit down and converse with the partner about recent trips and experiences.
roles:wearer = unknown, partner = unknown
unrelated events:X000, X001, X002 · validation errors: 0

#	sub-goal	clip time	initiated_by	completion	audio_necessity	events	rationale
g1	Sit down and answer the partner's questions about the supermarket.	0:03.1–0:33.1	partner_request	speech	speech_only	S000 S001 S002	The partner asks the wearer to sit down and inquires about their trip to the supermarket in S000 to S002.
g2	Listen to the partner describe the numerous random stops on their train ride.	0:45.2–1:14.3	unknown	speech	speech_only	S003 S004 S005 S006 S007 S008 S009	The partner talks about the train being empty but having more stops than expected in S003 to S009.
g3	Listen to the partner's explanation of the train stops in tech hub areas.	1:14.3–1:46.6	unknown	speech	speech_only	S010 S011 S012 S013 S014 S015	The partner discusses how the train stops cater to tech workers in the South Bay and San Francisco in S010 to S015.
g4	Listen to the partner's description of the final leg of the trip to downtown and the weather.	1:48.3–2:30.5	unknown	speech	speech_only	S016 S017 S018 S019 S020 S021	The partner explains the route past the airport and praises the beautiful weather downtown in S016 to S021.

事件时间线(只用 speech + sound,不用 narration)

id	clip time	kind	speaker / object	text	partner direction · distance	source
S000	0:03.1–0:04.1	speech	partner	Come come sit down. topic: inviting someone to sit down	front-left · 2.02 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S001	0:10.6–0:12.5	speech	partner	How was the supermarket? topic: asking about the supermarket experience	front-left · 0.99 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S002	0:31.9–0:33.1	speech	partner	Oh, did you get any fruit? topic: asking about purchasing fruit	front-left · 1.25 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S003	0:45.2–0:49.6	speech	partner	Well, the nice part is it's super empty. topic: commenting on trains being empty	front-left · 1.32 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S004	0:53.9–0:55.4	speech	partner	There's a lot of stops. topic: stating trains have many stops	front-left · 1.34 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S005	0:57.2–0:58.7	speech	partner	More stops than I remembered. topic: remembering fewer stops previously	front-left · 1.33 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S006	0:58.8–1:02.6	speech	partner	There's like really random locations. topic: describing stop locations as random	front-left · 1.35 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S007	1:03.0–1:04.9	speech	partner	Like like I was I was saying to my mom like. topic: beginning anecdote told to mother	front-left · 1.32 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S008	1:07.4–1:10.0	speech	partner	That's not a real place. That's a city in Texas. topic: questioning stop name relevance locally	front-left · 1.34 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S009	1:12.2–1:14.3	speech	partner	But they want to stop at that spot topic: mentioning train stops at San Antonio	front-left · 1.34 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S010	1:14.3–1:19.5	speech	partner	I guess like well I think it used to be like a big hub for people working in tech. topic: speculating stop's origin as tech hub	front-left · 1.34 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S011	1:20.8–1:23.3	speech	partner	Like to the South Bay or the San Francisco. topic: naming related tech destinations	front-left · 1.36 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S012	1:23.5–1:26.8	speech	partner	But it's like there's so all these little in-between stops. topic: criticizing numerous intermediate stops	front-left · 1.38 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S013	1:28.5–1:32.3	speech	partner	It's like there should just be you know the San Jose stop topic: suggesting San Jose as a major stop	front-left · 1.38 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S014	1:33.6–1:35.2	speech	partner	And then the Palo Alto stopped topic: suggesting Palo Alto as a major stop	front-left · 1.39 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S015	1:41.3–1:46.6	speech	partner	But I don't know they want to like put all these little ones in and then some trains stopped at all of them. topic: contrasting proposal with reality of many stops	front-left · 1.4 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S016	1:48.3–1:52.0	speech	partner	Like the train I was on it was stopping really early on. topic: describing own train making early stops	front-left · 1.41 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S017	1:52.5–1:56.8	speech	partner	But then once we got like kind of past the airport it was just a straight shot to downtown. topic: describing own train running express past airport	front-left · 1.41 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S018	2:01.5–2:03.9	speech	partner	It's all like equine distant because we're so like in the middle. topic: describing current location as equidistant	front-left · 1.4 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S019	2:05.6–2:09.9	speech	partner	We're just as close to downtown as we are to the airport essentially. topic: clarifying equidistant location relative to downtown and airport	front-left · 1.4 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S020	2:14.3–2:15.5	speech	partner	Yeah downtown was nice. topic: commenting positively on downtown area	front-left · 1.43 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
S021	2:26.9–2:30.5	speech	partner	Gorgeous blue skies. You know you you don't get that all the time. topic: commenting on pleasant weather and blue skies	front-left · 1.44 m	savvy-bench speech (QA-selected sentence, human-reviewed QA; speaker = is_me_speak)
X000	2:29.3–2:31.5	sound	dog	Dog walking on hard surface	–	ours-legacy sound (unverified, model/hand made)
X001	2:33.7–2:36.1	sound	dog	Dog walking on hard surface	–	ours-legacy sound (unverified, model/hand made)
X002	2:47.2–2:48.1	sound	dog	Dog walking on hard surface	–	ours-legacy sound (unverified, model/hand made)

3. 下一步

人工核验三个草稿的 sub-goal 边界、audio_necessity、roles;定 activity_type 是否够当「官方 step」。
通过后写 pyramid2 的 aea adapter(task_segment = goal,task_step = sub-goal,事件 = speech/sound),跑 Gemini 4 / Qwen3-Omni。
扩到 ≥10 句的 34 条录像(pairs 两侧都有标注的 30 对可做「谁说的」题)。