Emotions
There seem to be no emotions, can you add it or how is it used?
You can control the emotion by giving a reference with that emotion. We are actively working on a more explicit control. Until then this is the way to go.
örnek kullanım gösterme şansın var mı
How can we give reference to the emotion? Do we have to upload the emotional audio in speaker_wav?
@eysharaazia
@PratikHackTR
From the documentation:
https://tts.readthedocs.io/en/latest/
There seems to be a parameter you can add:
tts.tts_to_file(text="This is a test.", file_path=OUTPUT_PATH, emotion="Happy", speed=1.5)
xtts v2 does not appear to support emotions on all languages. I tested it with german and it completely ignores any emotions through conditioning.
It appears to be mostly an english feature
@nokked but it does not seem to work. But yeah if you provide the audio incorporated with emotion in speaker_wav parameter then you can see its affect in the audio.
When I output a text via the xtts_v2 model as a voice clone, the end of the text is sometimes abruptly cut off. For example, the last letter of the last word is missing. As a workaraound, I helped myself by adding a phrase like “but” at the end of the last sentence. Of course it's very unpleasant to have to edit out this "but" manually in audio editing afterwards.