What does Common Voice do?

Common Voice works in Speech-to-Text / Transcription, Datasets & Training Data — Mozilla's open, crowdsourced multilingual speech dataset for training speech recognition..

Is Common Voice open source?

Yes, Common Voice is open source (MPL-2.0 license). The source is on https://github.com/common-voice/common-voice.

Where is Common Voice based?

Common Voice is headquartered in United States.

Common Voice

Q: What is Common Voice?

Common Voice is a voice AI product — Mozilla's open, crowdsourced multilingual speech dataset for training speech recognition.

Mozilla's open, crowdsourced multilingual speech dataset for training speech recognition.

Speech-to-Text / Transcription Datasets & Training Data

Visit website

About Common Voice

Common Voice is Mozilla's open, crowdsourced initiative to build the world's largest publicly available multilingual dataset of human voices — so anyone can train speech-enabled AI. Volunteers donate recordings by reading short sentences and validate others' clips, producing a growing, openly licensed corpus of audio with transcripts across 100+ languages.

What's in it

Millions of validated voice clips with matching transcripts
100+ languages, including many low-resource ones
Optional, anonymized demographic metadata (age, gender, accent)
Released under CC0 (public domain) for unrestricted use

Who Is It For?

ML researchers and engineers training speech recognition (ASR/STT) and speech models
Teams building voice tech for under-served and low-resource languages
Anyone who wants to contribute their voice to open speech AI

Use Cases

Training and benchmarking speech-to-text and speech models
Improving language and accent coverage for voice AI
Academic research on speech and language

Open Source

Both the dataset (CC0) and the Common Voice platform code are open. Contribute clips at commonvoice.mozilla.org or the code on GitHub.

Help build open speech data for everyone — Mozilla Common Voice.