Researchers from Shanghai Jiao Tong University and the University of Cambridge have developed F5-TTS, a fully non-autoregressive text-to-speech system based on flow matching with Diffusion Transformer (DiT). In other words, it just might be the best AI text-to-speech generator yet.
This model was trained on a public 100,000 hour multilingual dataset, and as you can hear in the demos, Fairytaler Fakes Fluent and Faithful speech with Flow matching (F5-TTS) exhibits highly natural, expressive zero-shot ability, seamless code-switching capability, as well as speed control efficiency.
- Outdoor-Proof Speaker: Portable design with IPX7 waterproof protection to safeguard against splashes, waves, and water vapor. Get incredible sounds at...
- 24H Non-Stop Music: With Anker's world-renowned power management technology and a 5,200mAh Li-ion battery, the soundcore 2 speaker delivers a full day...
- Powerful Sound: The speaker features 12W power with enhanced bass from dual neodymium drivers. An advanced digital signal processor ensures pounding...
Without requiring complex designs such as duration model, text encoder, and phoneme alignment, the text input is simply padded with filler tokens to the same length as input speech, and then the denoising is performed for speech generation, which was originally proved feasible by E2 TTS,” said the researchers.
