liva-ai/couples-call-samples

Couples Call Sample Data

Technical Analysis

Measure	Value
Sample rate	48 kHz
Bit depth	24-bit PCM
File format	WAV (stereo, one person per channel)
SNR	mean 17.0 dB, median 17.1 dB (median speech level vs median level when nobody speaks)
Median speech RMS	-29.3 dBFS
Average speech ratio	0.24 per person
Spectral centroid	about 0.78 kHz
Full-band audio	82 % of hours carry speech above 20 kHz

Distribution (share of hours):

Measure	Low	Mid-low	Mid-high	High
SNR	< 10 dB: 14 %	10-20 dB: 50 %	20-30 dB: 31 %	>= 30 dB: 5 %
DNSMOS OVRL	< 2.0: 14 %	2.0-2.5: 34 %	2.5-3.0: 47 %	>= 3.0: 2 %
Bleed (other person's voice in the track)	< 5 %: 16 %	5-20 %: 4 %	20-40 %: 24 %	>= 40 %: 41 % (other 15 %: not two-person calls)
Bandwidth	< 8 kHz: 1 %	8-16 kHz: 8 %	16-20 kHz: 7 %	>= 20 kHz: 82 %
Speech ratio	< 0.1: 6 %	0.1-0.2: 27 %	0.2-0.3: 39 %	>= 0.3: 27 %

Dataset Overview

Item	Value
Average call duration	17.3 min (median 11.8 min) for two-person calls
Relationship	Romantic partners (couples app)
Erotic content	0.5 % of two-person conversation time
Code-switching	Yes, in 24 % of calls, mostly brief (4 % of calls have a minute or more in another language)

Speakers per Conversation

Share of recorded hours.

Speakers	Share of hours
1	16.0 %
2	83.7 %
3+	0.4 %

Conversation Categories

Share of hours of two-person conversations, labelled by a language model from the transcripts.

Questions

Question category	Share of hours
Deeper emotional questions	5.8 %
Shared memories	2.1 %
Spicy / intimate topics	1.5 %
Relationship boundaries	1.2 %
Future plans	0.7 %
Funny hypotheticals	0.4 %

Games

Game	Share of hours
Draw & Guess	15.8 %
Hot Take	9.3 %
Mini Golf	7.5 %
Same Wavelength	7.1 %
Word games	7.1 %
Connect 4	6.8 %
Who's More Likely	6.6 %
Cooperative puzzles / mysteries	2.3 %
Other games	2.0 %

Other

Activity	Share of hours
Free conversation	16.9 %
Setup / technical	6.9 %

Category Subsets

Five example excerpts (up to 2 minutes each, from different calls) for each question category, and five from different games; pick a subset in the dataset viewer.

Subset	Category
funny_hypotheticals	Funny hypotheticals
shared_memories	Shared memories
future_plans	Future plans
relationship_boundaries	Relationship boundaries
deeper_emotional_questions	Deeper emotional questions
spicy_intimate_topics	Spicy / intimate topics
games	Games (one each: Draw & Guess, Hot Take, Same Wavelength, Connect 4, Mini Golf)

Accent Subsets

Excerpts of up to 2 minutes from different couples; the accent is approximated from the speaker's location. accent_speaker says which channel carries it; each speaker's location is in the metadata.

Subset	Accent	Samples
accent_eastern_european_influenced_english	Eastern European influenced English	1
accent_general_american	General American	2
accent_latin_american_influenced_english	Latin American influenced English	1
accent_southern_american_english	Southern American English	2
accent_southern_european_influenced_english	Southern European influenced English	2
accent_western_european_influenced_english	Western European influenced English	2

Languages

By the language detected on each track.

Language	Share of hours
English	84.2 %
Spanish	4.8 %
German	2.8 %
Finnish	1.8 %
Italian	1.4 %
Filipino	1.2 %
Afrikaans	1.1 %
Swedish	1.0 %
Portuguese	0.8 %
Other	0.9 %

Accents

Estimated from where each speaker was when recording (IP location), not labelled from the audio.

English

Accent	Share of hours
General American	27.5 %
Southern American English	22.0 %
Northeastern American English	10.9 %
Western European influenced English	8.6 %
Unknown	5.6 %
Australian English	4.1 %
Southern European influenced English	4.1 %
Southern British English	4.1 %
Canadian English	3.9 %
Northern British English	3.6 %
East Asian influenced English	1.8 %
Middle Eastern influenced English	0.9 %
Maritime Southeast Asian English	0.6 %
Irish English	0.5 %
Southern African English	0.4 %
Latin American influenced English	0.4 %
Caribbean English	0.4 %
Mainland Southeast Asian influenced English	0.4 %
South Asian English	0.3 %
West African English	0.1 %