WorldSpeech: 65k hours of multilingual speech
A multilingual ASR corpus with 65k hours of human transcripts for 127 language variants.
- Automatic speech recognition
- Text → speech
- Audio classification
- Afrikaans
- Amharic
- Arabic
- +9
Hugging Face datasets with plain-language descriptions: what's inside and what models to train on them. 4,788 in the catalog so far, with new ones added every hour.
Answered by Gemini 4 · the chat is kept only in this tab
A multilingual ASR corpus with 65k hours of human transcripts for 127 language variants.
Large-scale collection of audio recordings, texts, and documents in rare Indian languages and dialects for ASR systems.
A large internal speech corpus in the Moore language with French translations and metadata for ASR and TTS tasks.
Large multilingual audio dataset for speech recognition converted from Mozilla Common Voice version 17.
Nearly 16 hours of Telugu audio with detailed transcription, timecodes, and speech feature markup.
Unofficial export of the 22nd version of Mozilla's open speech dataset for speech recognition tasks.
Audio recordings of everyday sentences in Irish English.
A small sample of two-way conversations with separated speaker tracks for ASR.
Audio and text for ASR and TTS across dozens of African languages—a large open speech corpus from Google.
Around 893,000 chapters from 96,000 Russian audiobooks with audio files and descriptions.
Read Khmer speech based on Tipitaka texts for ASR and TTS tasks with manual access moderation.
Large-scale extraction of texts, images, audio, and video from open datasets in a unified parquet format.
271 audio files featuring acoustic impulse responses for speech augmentation and reverberation simulation.
16 kHz Russian speech audio recordings with manual annotations from the crowdsourced part of the Golos corpus for ASR tasks.
A large heterogeneous English speech corpus with transcriptions and ready-made subsets for training ASR models.
An unofficial copy of the open Mozilla Common Voice 16.0 speech corpus for ASR tasks.
Large weakly-supervised ASR corpus covering five Arabic dialects.
Female ASMR voice recordings in AAC format with Whisper word-level alignment and non-speech sound tags for ASR and generation tasks.
Large English audiobook speech and text corpus (10m to 100m records) optimized for streaming.
Over 133,000 hours of voiced books across 86 languages for training speech recognition and synthesis systems.
A large-scale audio dataset and transcripts of Legislative Council meetings in Cantonese, featuring raw and segmented recordings.
Speech corpus of 26,867 audio readings by 24 non-native English speakers with manual phoneme labeling for ASR and pronunciation analysis.
Nearly 12,000 hours of Russian speech with consensus transcriptions from seven ASR models and added punctuation.
Wideband Arabic speech in segments for training ASR and TTS models.