Dispatch
Hume AI Open-Sources TADA: TTS Model with Zero Content Hallucinations, 0.09 RTF, Multilingual Support
Hume AI released TADA (Text-Audio Dual Alignment), its first open-source TTS model, on March 10. TADA generates text and audio in one synchronized stream, eliminating the text-audio desync that causes content hallucinations in prior LLM-based TTS systems — reporting zero hallucinations across 1,000+ test samples. At 0.09 RTF it runs 11x faster than real-time and 5x faster than comparable LLM TTS. Two models released: tada-1b (English) and tada-3b-ml (multilingual: Arabic, Chinese, German, Spanish, French, Italian, Japanese, Polish, Portuguese).
Source
↳ Follow the thread