Unify inference scripts (addresses #21)

[Ryu1845]

Unify inference scripts into a single CLI.

[Ryu1845]

Finished converting tortoise-tts.py. Needs to handle the other scripts now.

[152334H]

My apologies for adding the --voicefixer arg in the interim.

[Ryu1845]

Thanks for telling me :wink:

[Ryu1845]

I'll try to finish tomorrow. I can't promise anything though.

[152334H]

sure no worries

[Ryu1845]

Only need to add voicefixer and it's done I think.

[Ryu1845]

Nvm I need to merge app.py and change the readme too

[152334H]

🥹

[Ryu1845]

For some reason the webui is bugging when I call it as a function... I should probably make a proper webui but I can't be arsed with making frontend tbh

[Ryu1845]

I'll open for review without the webui option for now, it's not really necessary tbh.

[Ryu1845]

Oh wait I forgot to add the voicefixer toggle in the end :laughing:

[Ryu1845]

There, done (exclude launching webui from the cli)

[Ryu1845]

@152334H Could you review?

[152334H]

I would say it's a bit unintuitive that the script shows no output at all when no input text is given, and I'm adding a basic print statement to note when text is being read from stdin.

[xenotropic]

One thing that seemingly did not make it in this refactor / merge is the ability to control how text is broken up -- read.py would separate on pipe symbols | . That allowed for the user to choose how chunks are broken up rather than necessarily using --text_split , which often cuts in the middle of sentences, leading to choppier outputs (a pause and "sentence start" emphasis mid-sentence). Feature request; thanks for considering.