Speech & Audio datasets

Speech and audio datasets for automatic speech recognition, speech synthesis, speaker analysis and voice interfaces. Listings cover scripted and spontaneous speech, call-centre conversations and far-field recordings across a wide range of languages and accents.

The catalogue lists 1,727 speech & audio datasets. Most listings are audio, text and multimodal data, typically used for conversational AI, speech recognition and LLM pre-training. Where providers state a volume in hours, it ranges from 1 hours to 10M hours.