Is there a simple way to do inference? Without gradio or jupyter. Looking to run this on a batch of files.
Also, there's one thing i don't understand: i'm expecting to have 3 basic parameters:
- original audio (the one that will be used for cloning)
- original audio transcript
- target text to convert to speech
Now in the jupyter file i see the target text has to start with something from the original audio. I'm expecting to use a separate target that doesn't contain parts of the original audio.
The quality is absolutely great, but it always outputs a piece of the original audio at the start of each generated speech. Also, generation is really slow (on a 4090, and it eats up all the VRAM)
Is there a simple way to do inference? Without gradio or jupyter. Looking to run this on a batch of files.
Also, there's one thing i don't understand: i'm expecting to have 3 basic parameters:
Now in the jupyter file i see the target text has to start with something from the original audio. I'm expecting to use a separate target that doesn't contain parts of the original audio.
The quality is absolutely great, but it always outputs a piece of the original audio at the start of each generated speech. Also, generation is really slow (on a 4090, and it eats up all the VRAM)