James Betker · 2022 · slow diffusion
Deliberately slow. An autoregressive GPT-like stage plus a diffusion decoder, famous for cloning from short references if you are willing to wait. The clips are official repo examples: an LJSpeech fine-tune, and the ‘lax’ pair from the NaturalSpeech comparison (not a celebrity clone).
Quality by burning latency on purpose.
Vendor / paper demo
Tortoise TTS (NaturalSpeech comparison, ‘lax’) · Official tortoise-tts examples/naturalspeech_comparison/lax/tortoise.mp3, the same LJSpeech sentence as NaturalSpeech’s ‘lax discipline’ demo. Celebrity-clone examples in the repo are not hosted here.
The lax discipline maintained in Newgate was still further deteriorated by the presence of two other classes of prisoners who ought never to have been inmates of such a jail.
Vendor / paper demo
Tortoise TTS (LJSpeech fine-tune) · Official tortoise-tts examples/finetuned/lj/2.mp3.
Vendor demo: LJSpeech fine-tune clip from tortoise-tts examples/finetuned/lj/2.mp3. Exact spoken text is not printed in the README; listen to the file.