StyleTTS2 icon indicating copy to clipboard operation
StyleTTS2 copied to clipboard

Audio Length Customization

Open zzryn opened this issue 11 months ago • 1 comments

I want to find a way to increase the 300-second limit. Is there any way to do that?

zzryn avatar Jan 07 '25 02:01 zzryn

The 300-second limit can be bypassed by addressing the underlying constraint of BERT's maximum token length of 512. You can tokenize your text into sentences and process them one by one. For each chunk, generate the corresponding audio and then stitch the audio together to create a seamless output of any desired length.

UmerrAhsan avatar Jan 07 '25 09:01 UmerrAhsan