Speech model evaluation benchmark (SUPERB)
English audio and text data for evaluating universal speech processing models.
- Automatic speech recognition
- Audio classification
- English
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
English audio and text data for evaluating universal speech processing models.
Large Russian speech corpus from webinars (over 315 thousand fragments) with texts for ASR and TTS tasks.
31 hours of transcribed studio audio from 120 speakers of British and Irish dialects for TTS.
Room impulse responses at 32 kHz for training ASR models to operate in echo conditions.
An audio dataset for medical speech recognition across multiple languages.
An audio dataset of domestic dinner parties for speech separation and recognition tasks in real-world conditions.
English spontaneous speech audio recordings and texts for cognitive impairment assessment.
Anime character audio clips of Kasugano Sora for speech generation training.
Cleaned audio recordings and texts in Vietnamese from 152 speakers for speech synthesis and recognition tasks.
Large corpus of audio and transcription pairs in Japanese from adult games for ASR and TTS tasks.
An audio corpus of everyday dialogues in 21 languages for speech recognition and speaker diarization.
Аудиодатасет реплик из игры AzurLane на японском и китайском языках для дообучения GPT-SoVITS.
Lightweight Piper (VITS) speech synthesis model checkpoints with a British accent for real-time CPU text-to-speech.
Annotated English video calls from NVIDIA for evaluating conversational multimodal agents.
19 hours of paired male and female voice recordings for training lightweight TTS models.
Metadata of nearly 660k Suno AI songs with multilingual descriptions and audio modality.
Around 5.3 hours of Sanskrit shloka recitation audio by a single speaker for TTS training.
Audio recordings with spectrograms and metadata for classifying rare and traditional instruments of China.
Parallel corpus of texts and audio recordings in English and six East African languages.
Thousands of Japanese anime speech audio fragments from visual novels with transcripts for ASR.
Unprepared telephone conversations in five languages with annotation for speaker voice separation.
Large collection of audio recordings with transcriptions in Brazilian Portuguese for training speech recognition systems.
Massive Audio-FLAN corpus in English and Chinese for speech, music, and sound understanding and generation.
A benchmark dataset of 74,858 Vietnamese audio recordings with IPA transcriptions for speech synthesis and recognition.