Govar Data
Unscripted English speech data for voice AI
License natural, multi-speaker English conversations from mostly non-native speakers. Participants recorded remotely on their own devices, with a separate track for each speaker.
Request a sample2,100+ hours of estimated speech
150–500 hours added per month
Dataset specifications
- Speech
- Unscripted, multi-speaker English conversations.
- Audio
- Separate speaker tracks recorded on participants’ own devices.
- Metadata
- Country, English proficiency and native language.
- Transcripts
- Available by arrangement.
What you can evaluate
- Accented English ASR
- Evaluate recognition on natural conversations with mostly non-native speakers.
- Diarization and separation
- Use individual speaker tracks to evaluate who spoke and when.
- Turn-taking
- Study conversational timing for speech-to-speech systems.
Consent and licensing
The dataset is available for licensing to voice AI labs with participant consent. Ask us for sample access, consent documentation and licensing terms.
Custom collection
We can arrange conversational data collection in other languages or on specific topics. Tell us the language, topic and recording setup you need.
Discuss a custom collection