New TTS model to play with – ViiTorVoice

https://github.com/viitor-ai/viitor-voice-nar

图片

Whilst it is not the only benchmark which matters, but having such low WER for both English and Chinese is pretty impressive. About one word wrong out of every 100 words — which in TTS evaluation is excellent and implies the synthesized speech is highly intelligible.

Related posts:

  1. Will #Google release the trillion-parameter model for download? Or will it keep it guarded like #GPT3 and plan to make a lot of money from it? #AI venturebeat.com/2021/01/12/goo…
  2. RT @TrungTPhan: When a startup claims it has “AI” or “ML” and then you take a look at the actual tech pic.twitter.com/2acvwXiJpy
  3. Assuming that #AI will one day be unbiased and ethical, what will we do with humans not strictly abiding to it? venturebeat.com/2020/12/10/tim…
  4. “Less than one”-shot (LO-shot) learning. Next gen #AI will need less data and processing power. technologyreview.com/2020/10/16/101…