--- pretty_name: "Indian-language calls between people and an AI voice agent" language: - as - bn - en - gu - hi - kn - ml - mr - or - pa - ta - te - ur language_details: as-IN, bn-IN, en-IN, gu-IN, hi-IN, kn-IN, ml-IN, mr-IN, or-IN, pa-IN, ta-IN, te-IN, ur-IN license: other license_name: lokah-custom license_link: https://kenpathlabs.com/lokah/licensing task_categories: - audio-to-audio - automatic-speech-recognition size_categories: - n<1K tags: - Assamese speech dataset - Bengali speech dataset - Indian English speech dataset - Gujarati speech dataset - Hindi speech dataset - Kannada speech dataset --- # Indian-language calls between people and an AI voice agent > Phone calls between a person and an AI voice agent in thirteen Indian languages, on recruitment, telecom, healthcare, e-commerce and insurance, each voice on its own channel at 48 kHz. A sample release: 27 calls, 100 minutes, with verbatim transcripts of both sides. Recorded to order in your languages, domains and scenarios, at any volume. `LK-SP-MUL-001` · [Get a quote](https://kenpathlabs.com/lokah/datasets/ai-voice-agent-calls#contact) · [Record as JSON](https://kenpathlabs.com/api/lokah/datasets/ai-voice-agent-calls) · [Croissant](https://kenpathlabs.com/lokah/datasets/ai-voice-agent-calls/croissant.json) Phone calls between a person and an AI voice agent in thirteen Indian languages, on recruitment, telecom, healthcare, e-commerce and insurance, each voice on its own channel at 48 kHz. A sample release: 27 calls, 100 minutes, with verbatim transcripts of both sides. Recorded to order in your languages, domains and scenarios, at any volume. Measured from 27 sample conversations (100 minutes): 48 kHz, 16-bit FLAC, one two-channel file per conversation. Across the whole set 16% of transcript words are English written in Latin script and 24% of the time is silence; the sample conversations hold 683 turns. 21 callers (5 male, 16 female) and one AI voice agent (female voice). Every recording has a complete, segment-level transcript. Personal data: redacted. Licence: custom, quoted per use. ## Specification | Field | Value | Note | | --- | --- | --- | | id | LK-SP-MUL-001 | | | type | speech · conversational · call centre | | | language | অসমীয়া · Assamese · as-IN / বাংলা · Bengali · bn-IN / English · Indian English · en-IN / ગુજરાતી · Gujarati · gu-IN / हिन्दी · Hindi · hi-IN / ಕನ್ನಡ · Kannada · kn-IN / മലയാളം · Malayalam · ml-IN / मराठी · Marathi · mr-IN / ଓଡ଼ିଆ · Odia · or-IN / ਪੰਜਾਬੀ · Punjabi · pa-IN / தமிழ் · Tamil · ta-IN / తెలుగు · Telugu · te-IN / اردو · Urdu · ur-IN | | | hours | 2 h | | | channels | Two channels, one per speaker | | | files | 27 files · 27 conversations | counted across the sample release | | layouts | 27 two-channel | counted across the sample release | | audio | FLAC · 48 kHz · 16-bit | measured across the sample release | | bandwidth | wideband (8 kHz) | | | snr | 38.1 dB median | across the sample release, on the caller's channel | | release | sample release | the full collection is recorded to order | | transcript | time-aligned by segment · Bengali-Assamese and Bengali and Latin and Gujarati and Devanagari and Kannada and Malayalam and Odia and Gurmukhi and Tamil and Telugu and Perso-Arabic script · delivered as JSON | | | speakers | 21 callers and one AI voice agent (female synthetic voice) | | | pii | Numbers of 4 or more digits, digit strings spoken as words, and email addresses are masked as [pii] in text and replaced by a tone in audio. First names are not masked. | | | source | Recorded to order with an AI voice agent | | | review | Every recording has a complete, segment-level transcript | | | personal data | Redacted | | | licence | Custom | quoted per use | ## What it is good for | Task | Fit | Why | | --- | --- | --- | | Speech recognition | yes | Time-aligned transcripts in native script, English kept as spoken. 38.1 dB median SNR across the full set, wideband (44.3 dB in the sample conversations). | | Full-duplex speech to speech | yes | Each speaker on a separate channel, so overlap, backchannels and turn timing survive. Moshi and PersonaPlex train on exactly this layout. | | Turn-taking and endpointing | yes | Gaps and overlaps at every change of speaker are measured from the two channels; see the profile. | | Voice agents for support | yes | Caller and agent turns across support and recruitment scenarios, with the agent on its own channel. | | Speaker diarisation | yes | Speaker-attributed segments across full conversations. | | Text to speech | partly | 38.1 dB median SNR across the full set, wideband (44.3 dB in the sample conversations): clean and wideband enough for conversational prosody data, though not a studio voice. | ## Conversation profile Across every conversation in the dataset (27 conversations, median 3.73 minutes), as measured by the delivery. | Measure | Value | | --- | --- | | Talk time, most-talking speaker / the other | 60% / 40% | | Silence | 24% | | Overlapping speech | 2% | | English words, written in Latin script | 16% | | Turns a minute | 6.39 | | Speaking rate | 120 words a minute | ### Turn-taking, measured from the two channels across the set | Per minute | This dataset | Fisher corpus | | --- | --- | --- | | Inter-pausal units | 18.46 | 21.6 | | Pauses | 9.28 | 7 | | Gaps | 6.14 | 7.5 | | Overlaps | 0.87 | 6.5 | | Backchannels | 1.76 | not reported | Floor-transfer offset: median +1.36 s, 10th to 90th percentile -0.16 s to +2.25 s. Channel isolation: 60 dB. ## Audio quality Frame RMS at 20 ms on the channels mixed to mono. Noise floor is the 10th percentile, speech level the 90th; SNR is their difference. Bandwidth is the highest frequency at which speech still rises 6 dB above the recording's own noise spectrum. | Measure | Value | Reading | | --- | --- | --- | | SNR across every file of the set | 38.1 dB median | caller's channel | | Speech above noise floor (SNR), sample conversations | 44.3 dB (20.7 to 62.9) | clean | | Noise floor | -62.9 dBFS | | | Effective bandwidth | 8.0 kHz | wideband (8 kHz) across the full set | | Clipping | 0.002% of samples | | | DNSMOS P.835 (1 to 5) | background 4.01, speech 3.54, overall 3.24 | listener-rated quality, estimated | ## Samples 26 public excerpts, 36 to 89 seconds each, from different conversations in the dataset. - Sample 1, Customer service: mobile network complaint (Odia), from a 4-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--1.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--1.segments.json) - Sample 2, Customer service: two-wheeler insurance claim (Gujarati), from a 4-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--2.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--2.segments.json) - Sample 3, Customer service: e-commerce return and refund (Bengali), from a 4-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--3.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--3.segments.json) - Sample 4, Recruitment: phone screening (Tamil), from a 4-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--4.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--4.segments.json) - Sample 5, Recruitment: phone screening (Kannada), from a 4-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--5.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--5.segments.json) - Sample 6, Customer service: clinic appointment booking (Assamese), from a 3-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--6.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--6.segments.json) - Sample 7, Recruitment: phone screening (Telugu), from a 4-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--7.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--7.segments.json) - Sample 8, Customer service: e-commerce return and refund (Malayalam), from a 5-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--8.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--8.segments.json) - Sample 9, Customer service: e-commerce return and refund (Hindi), from a 4-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--9.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--9.segments.json) - Sample 10, Customer service: e-commerce return and refund (Indian English), from a 3-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--10.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--10.segments.json) - Sample 11, Customer service: two-wheeler insurance claim (Urdu), from a 2-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--11.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--11.segments.json) - Sample 12, Recruitment: customer-support role interview (Marathi), from a 4-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--12.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--12.segments.json) - Sample 13, Customer service: clinic appointment booking (Punjabi), from a 2-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--13.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--13.segments.json) - Sample 14, Customer service: clinic appointment booking (Bengali), from a 3-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--14.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--14.segments.json) - Sample 15, Customer service: e-commerce return and refund (Tamil), from a 2-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--15.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--15.segments.json) - Sample 16, Customer service: clinic appointment booking (Tamil), from a 3-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--16.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--16.segments.json) - Sample 17, Recruitment: customer-support role interview (Assamese), from a 5-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--17.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--17.segments.json) - Sample 18, Customer service: mobile network complaint (Assamese), from a 4-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--18.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--18.segments.json) - Sample 19, Customer service: clinic appointment booking (Gujarati), from a 4-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--19.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--19.segments.json) - Sample 20, Customer service: clinic appointment booking (Kannada), from a 4-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--20.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--20.segments.json) - Sample 21, Customer service: mobile network complaint (Hindi), from a 3-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--21.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--21.segments.json) - Sample 22, Customer service: mobile network complaint (Telugu), from a 4-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--22.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--22.segments.json) - Sample 23, Recruitment: customer-support role interview (Telugu), from a 6-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--23.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--23.segments.json) - Sample 24, Recruitment: phone screening (Marathi), from a 5-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--24.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--24.segments.json) - Sample 25, Customer service: mobile network complaint (Punjabi), from a 4-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--25.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--25.segments.json) - Sample 26, Recruitment: phone screening (Indian English), from a 3-minute conversation: [audio](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--26.m4a) · [segments](https://kenpathlabs.com/lokah-samples/ai-voice-agent-calls--26.segments.json) ### Sample 1 transcript Customer service: mobile network complaint (Odia), starting at 1:01. Stereo: left channel is the caller, right channel is the ai voice agent. Audio sha256 `edc33d0ba940929d99e06b151e3fb44d8e8c87e26f1b708588ee556ff4ec6ae3`. | Start | Speaker | Text | | --- | --- | --- | | 0:00 | AI voice agent | ଟିକେ ବିସ୍ତାରରେ କହିପାରିବେ କି? କେଉଁ ସବୁ ଆପ୍ସରେ ଆପଣଙ୍କୁ ଅସୁବିଧା ହେଉଛି? | | 0:10 | Caller | ଏଇଟା ମୋର ଏଇ ତିନି ଚାରି ଦିନ ହବ ହେଉଛି। ଆଉ ଟିକେ YouTube ଚଲେଇଲା ବେଳେ video ଅଳ୍ପ ଚାଲୁଛି, ଅଳ୍ପ ତା'ପରେ network slow ହଉଛି, ଗୋଲ ଗୋଲ ଘୂରି ଘୂରିକି ବହୁତ slow ହଉଛି। | | 0:27 | AI voice agent | ମୁଁ ବୁଝିପାରୁଛି। YouTube ଚଲାଇବା ସମୟରେ ଅସୁବିଧା ହେଉଛି। ଥରେ ଆପଣଙ୍କ ମୋବାଇଲ୍ଟିକୁ ରିଷ୍ଟାର୍ଟ କରି ଦେଖିବେ କି? ଏହାଦ୍ୱାରା ନେଟୱାର୍କ ସେଟିଂସ୍ ରିଫ୍ରେଶ ହୋଇଯାଏ। | | 0:38 | Caller | ହଁ, ସେଟା କରିକି ଦେଖିଥିଲି କିନ୍ତୁ ଏଟା | | 0:43 | AI voice agent | ଆଚ୍ଛା, ତେବେ ଗୋଟେ ଥର ଆପଣଙ୍କ ମୋବାଇଲର ନେଟୱାର୍କ ମୋଡ୍ ଟା ଥରେ ଚେକ୍ କରିବେ କି? ଏହା 4G କି 5G ରେ ସିଲେକ୍ଟ ଅଛି ନା 3G ରେ? | | 0:53 | Caller | ହେଉନାହିଁ। ଏଟା ଅଛି ମୋର five G network ରେ ହିଁ ଅଛି। | | 0:57 | AI voice agent | ହଁ, ଠିକ୍ ଅଛି। ଆଚ୍ଛା, ଆପଣ କେଉଁ ପ୍ଲାନରେ ଅଛନ୍ତି ଏବଂ ଦିନକୁ କେତେ ଡାଟା ୟୁଜ୍ କରନ୍ତି ଜାଣିପାରେ କି? | | 1:05 | Caller | ଏଇଟା ମୋର ମ-monthly ତ ଅଛି, ଅଂ, monthly ତ per day ତ ମୋର ଆସୁଛି two GB data। କିନ୍ତୁ two GB data ସରୁନି, network slow ଯୋଗୁଁ data ହିଁ ସରୁନି। | | 1:21 | AI voice agent | ଆଉ ଯେଉଁ ରିଚାର୍ଜ ଆପଣ କରିଥିଲେ, ତା'ର ଅମାଉଣ୍ଟ, କେଉଁ ତାରିଖରେ ଏବଂ କେଉଁ ପେମେଣ୍ଟ ଆପ୍ ଦ୍ୱାରା କରିଥିଲେ ଜାଣିପାରିବି କି? | ## Licence Licence: custom, quoted per use (training, evaluation or both; internal or commercial; exclusive or not). Delivered in the layout your training code reads: https://kenpathlabs.com/lokah/formats. --- Machine-readable: [llms.txt](https://kenpathlabs.com/llms.txt) · [JSON API](https://kenpathlabs.com/api/lokah/datasets) · [OpenAPI](https://kenpathlabs.com/lokah/openapi.json) · [MCP](https://kenpathlabs.com/api/lokah/mcp) · [for agents](https://kenpathlabs.com/lokah/for-agents)