In-car speech data, recorded by native speakers to your brief
We record wake words, voice commands, navigation and media requests, addresses and dialogue for automotive voice teams, with native speakers recruited to your quotas for age, gender, region and accent. In-car audio is one of the devices and environments your brief can specify, and every session is checked before delivery.
Your prompts, in the voices of your users
Your prompt list
Speakers record the prompts in your brief: the wake word, commands, navigation and media requests, and addresses. Monologues and two-person dialogue are possible too, for assistants that hold a conversation.
Speakers to your quotas
Native speakers are recruited to your quotas for age, gender, region and accent. Each passes an ID check and a microphone test, and we confirm their language or dialect before they record.
Checked, then transcribed
Every session goes through automatic checks (SNR, clipping, duration) and a human review. Sessions that miss spec are re-recorded, not shipped, and files that pass get time-aligned transcripts in your convention.
Seats, microphones and noise, set in your brief
What the brief can say about the cabin
| Speaker position | Driver, front passenger or rear seats |
|---|---|
| Microphone | Where it sits: headliner, sun visor, dashboard or a phone in a mount |
| Vehicle state | Parked with the engine off, idling or moving, with speakers in passenger seats while the car moves |
| Windows | Closed, partly open or fully open |
| Ventilation | Fan or air conditioning off, low or high |
| Media | Music or radio playing, and how loud |
| Cabin noise | Indicators, road noise, other people talking |
List the conditions that matter for your product and ask for each as a metadata field, so you can measure accuracy per condition. The feasibility assessment that comes with your quote covers each one.
Why cabin audio is hard for ASR
Cabin noise depends on speed, windows and ventilation. In CAVEMOVE, a 2025 preprint with acoustic measurements in four cars, normal-effort speech from the driver’s seat was 16.2 to 21.8 dB above the noise with the car standing still, the windows closed and the fan off. At 110 km/h with the front windows fully open it was 13.4 to 17.1 dB below it, and the ventilation alone, standing still with the windows closed, could bring it close to or below 0 dB.
Research corpora vary these conditions on purpose: AVICAR recorded each script idling, and at 35 and 55 mph with the windows open and closed. Test your model in the conditions your users drive in, and put them in the brief.
What a project can include
| Prompts | Wake word, commands, navigation and media requests, addresses, from your list |
|---|---|
| Speech | Sentence or prompt lists, monologues, or spontaneous two-person dialogue |
| Speakers | Native speakers recruited to your quotas for age, gender, region and accent |
| Devices | In-car, phone, headset or far-field, as your brief specifies |
| Transcripts | Time-aligned, human-verified, in your convention, verbatim or clean-read |
| QA | Automatic checks plus human review; sessions that miss spec are re-recorded, not shipped |
| Consent | Explicit commercial-use consent from every speaker, documented |
| Delivery | Audio, transcripts and per-speaker metadata in the layout your pipeline expects |
| Languages | Any of our 60 catalog languages and dialects, and others after a feasibility check |
Built for
- In-car voice assistants
- Wake word and command models for the cabin
- Voice control for navigation and media
- Testing ASR across cabin conditions
A wake word for the car?
A detector also needs phrases that sound like the wake word but should not trigger it, recorded by the same speakers. Our page on wake word data collection covers that brief, for the cabin or any other device.
Frequently asked questions
Can you record speech data in cars?
Yes. In-car audio is one of the recording environments a collection brief can specify, alongside phone, headset and far-field. Native speakers recruited to your quotas record your prompts, monologues or dialogue, and every session passes automatic checks and a human review before delivery. You get a feasibility assessment and a quote within one business day of sending your brief.
What can the brief specify about the cabin?
Where the speaker sits, where the microphone is, whether the car is parked, idling or moving (with the speaker in a passenger seat, never a driver reading prompts), the windows, the fan or air conditioning, music playing and other cabin noise. List the conditions that matter for your product and ask for each one as a metadata field, so you can measure accuracy per condition. The feasibility assessment that comes with your quote covers each condition you list.
Which languages can you record?
Any of the 60 languages and dialects in our catalog, and other languages on request. We confirm recruitment feasibility for each language before quoting.
How is in-car data collection priced?
Per project, based on the language, speaker count and quotas, speech type, devices and environments, transcription depth and timeline. You get a feasibility assessment and a quote within one business day of sending your brief.
Describe your cabin and your users
Share your prompts, languages and accents, speaker counts, seat and microphone positions and the cabin conditions you need. We’ll come back within one business day with a feasibility assessment and a quote.