Works exceptionally well! But can we implement chunking?

[MaTeZZ]

Thank you for this!
For long text, this is the most consistent zero-shot TTS. Works really well until about 2:40 of audio, so maybe up to 2 minute chunks?

[MaTeZZ]

I have added the chunking in my fork

[flesti]

stops after 40 sec for me

MaTeZZ implementation also cuts off or skips words or entire parts at the end of paragraphs.