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

feat(realtime): support sola algorithm #316

Merged
merged 1 commit into from
Apr 13, 2023
Merged

feat(realtime): support sola algorithm #316

merged 1 commit into from
Apr 13, 2023

Conversation

yxlllc
Copy link
Contributor

@yxlllc yxlllc commented Apr 13, 2023

I implemented a simple sola algorithm, see this issue for the reason.

Because the previous real-time processing codes are all based on numpy, I did not use torch to implement the gpu-accelerated version. After testing it, it is acceptable to use cpu calculations

I set the default sola search length to 384, you can consider turning it into an adjustable parameter of gui, I am not very good at writing this part of the code.

@Lordmau5
Copy link
Collaborator

Did a bit of testing on 2 models and it seems to work.

Around the same delay as well (~1s before, ~1s with these changes) so that's good.
(Obviously lower would be better but I'd assume we might need gpu-acceleration to improve that)

@34j
Copy link
Collaborator

34j commented Apr 13, 2023

Sorry for the inconvenience, but please run pre-commit.

git reset HEAD^
pip install pre-commit
pre-commit install
pre-commit run --all
git add .
git commit -m "feat(realtime): support sola algorithm"
git push --force

@34j 34j changed the title support sola algorithm feat(realtime): support sola algorithm Apr 13, 2023
@yxlllc
Copy link
Contributor Author

yxlllc commented Apr 13, 2023

OK, I have executed pre-compile

@34j 34j merged commit 0fcbf99 into voicepaw:main Apr 13, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Improve the sound quality of real-time voice changer with better stitching algorithm
3 participants