Datoric/tts-conversational-voice-20000h

TTS Voice Dataset 20,000 hours of high-fidelity 48kHz conversational audio across 30+ global, regional, and underrepresented languages, built for text-to-speech, voice cloning, and multilingual speech AI. This repository contains the full technical specification, annotation schema, and sample metadata files (Parquet). The production dataset is rights-cleared and delivered directly to buyers. Request access to see the full schema and get real audio samples. Overview… See the full description on the dataset page: https://huggingface.co/datasets/Datoric/tts-conversational-voice-20000h.