Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Audio truncated and wierd silences on Spanish XTTS #410

Open
MrE3gman opened this issue Mar 4, 2025 · 11 comments
Open

Audio truncated and wierd silences on Spanish XTTS #410

MrE3gman opened this issue Mar 4, 2025 · 11 comments

Comments

@MrE3gman
Copy link

MrE3gman commented Mar 4, 2025

Script Mode

Process Mode

  • Gradio GUI,

Operating System:

  • Windows

Describe the bug
Problem #262 reapeared
It was solved but using the latest version (git pull yesterday runing in Docker full sync on app directory on windows) it seems to have regressed to the original problem. The fix worked but seems to have been broken again

@DrewThomasson
Copy link
Owner

DrewThomasson commented Mar 4, 2025

Did you pull the latest docker as well?

@MrE3gman
Copy link
Author

MrE3gman commented Mar 4, 2025

Yes, I created the container from scratch yesterday following the git clone and syncing the directory with the provided compose

@ROBERT-MCDOWELL
Copy link
Collaborator

try to set the repetition_penalty to 2.4 in thumbnail "Fine tune parameters"

@MrE3gman
Copy link
Author

MrE3gman commented Mar 4, 2025

Repetition penalty at 2.4 still makes bad results, can't really say if it improves or makes it worse.

Also tried on native as well as docker, the results are the same.

@ROBERT-MCDOWELL
Copy link
Collaborator

ROBERT-MCDOWELL commented Mar 4, 2025

without any log and original text where the issues are happening I cannot do anything
also provide the xtts fine tuned parameters you set for the conversion

@MrE3gman
Copy link
Author

MrE3gman commented Mar 5, 2025

After some checking I have found thtat the problems come from punctuation in the ebook. All "..." and some other punctuations ("-" or "¿") produce wierd sounds that seem like a glitch.

In sentence splitting it seems that "¿" splits on the previous sentence wich may be a problem.

Converting 19.28%: : 2615/13565
Sentence: Sin embargo , este libro menciona cuánto grano es necesario para mantener varías guarniciones en el Imperio Final . ¿
Converting 19.28%: : 2616/13565
Sentence: Tenéis idea de cuánta comida necesita un ejército ? - Ahí tienes un buen argumento - dijo Clubs , asintiendo - . Normalmente ,
Converting 19.29%: : 2617/13565

I can share some audio if needed or run tests, but punctuation in spanish seems broken at some points. For parameters I used default on all, fresh from github with speed at 1.2 as the only change

@ROBERT-MCDOWELL
Copy link
Collaborator

could you provide the part of the text causing the issue?

@ROBERT-MCDOWELL
Copy link
Collaborator

should be fixed in the next update.

@ROBERT-MCDOWELL
Copy link
Collaborator

could you provide the text anyhow? I want to be sure it's fixed for spanish, so try with my dev version and share it to you to confirm it is fixed.

@MrE3gman
Copy link
Author

MrE3gman commented Mar 9, 2025

Sure, i use this as a test. It's the first few lines of a book containing "..." and "¿"

Prologo_Imperio.txt

@ROBERT-MCDOWELL
Copy link
Collaborator

is this audio better?

Prologo_Imperio.zip

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

3 participants