Mozilla Common Voice

by Mozilla Foundation

Crowd-sourced multilingual speech corpus — 30k+ hours across 130 languages.

TL;DR

Crowd-sourced multilingual speech corpus — 30k+ hours across 130 languages.

Best for multilingual ASR training, low-resource language baselines, accent/dialect coverage. Pricing: free.

Category
Open source
License
Stars
Last push
Pricing
free
Platforms
HuggingFace, Web

What it is

Mozilla Common Voice is the largest public crowd-sourced multilingual speech corpus. Volunteers read prompted sentences in their language; the validated subset is the training-grade slice. Versions ship every 6 months (cv-17.0 = 2024-03). License: CC0.

Best for: Multilingual ASR training, low-resource language baselines, accent/dialect coverage.
Watch out for: CC0 1.0 (public domain) · read-prompt speech · variable per-language quality (en/de/fr massive, many languages <50h) · validated vs. invalidated splits matter. Cite: Ardila et al., LREC 2020.

Install / use

from datasets import load_dataset; ds = load_dataset('mozilla-foundation/common_voice_17_0', 'en')

Features

Speaker diarizationNo
Word-level timestampsNo
Streaming / real-timeNo
Languages supported130
HIPAA eligibleNo

Mozilla Common Voice vs Whipscribe

FeatureMozilla Common VoiceWhipscribe
CategoryOpen sourceTranscription APIs
Pricingfreefree beta
Speaker diarizationNoYes
Word timestampsNoYes
StreamingNoNo
Languages13099
PlatformsHuggingFace, WebWeb, API, MCP

Alternatives to Mozilla Common Voice

Whipscribe is a managed faster-whisper + whisperX service. If you want transcripts without running infrastructure, paste a URL or drop a file in the form below — you'll have a transcript in seconds.