RnNoise v0.2

[pengzhendong]

No description provided.

[werman]

There are additional changes beside rnnoise update, could you add why change was made to the respective commits?

[pengzhendong]

There are additional changes beside rnnoise update, could you add why change was made to the respective commits?

They are all from the rnnoise update except for the google style and lv2 bug.

[XCroatoanX]

@pengzhendong
Hi, thank you very much for your PR!
Is there maybe a changelog how it will improve the noise suppresion?

[pengzhendong]

@pengzhendong Hi, thank you very much for your PR! Is there maybe a changelog how it will improve the noise suppresion?

Sorry, I don't know. You could refer to the original repository: https://github.com/xiph/rnnoise

[DmitriyFrogo]

how it will improve the noise suppresion?

@XCroatoanX
All time using/testing this plugin, I encountered some artifacts when somebody singing/screaming, also it has issues with sounds "ss", "sh", "ch"
In new version, most of them is gone

For personal testing purposes I created one file with part of song and processed it with same settings, but one from old RNNoise 1.03 and second from RNNoise + this PR
This song not DMCA-free, so I uploaded it here for you:
https://pixeldrain.com/l/TCmpJ4xa

The difference is obvious in first 5-10 seconds

[XCroatoanX]

The difference is obvious in first 5-10 seconds

Wow, thank you very much, the difference is indeed day and night, can’t wait to have the update on Arch!

[DmitriyFrogo]

can’t wait to have the update on Arch!

If you can't wait, like me, maybe you can build it for yourself like I did, by forking https://github.com/pengzhendong/noise-suppression-for-voice and using Github Actions workflow

[pengzhendong]

can’t wait to have the update on Arch!

If you can't wait, like me, maybe you can build it for yourself like I did, by forking https://github.com/pengzhendong/noise-suppression-for-voice and using Github Actions workflow

I built the LADSPA plugin with MSVC using a GitHub Actions workflow, but it failed. However, it succeeded on my local machine. When I upgraded the MSVC version from 19.37.32824.0 to 19.39.33523.0, I reproduced this issue.

https://github.com/pengzhendong/noise-suppression-for-voice/actions/runs/9121513325/job/25080780130

[DmitriyFrogo]

@pengzhendong
Strange thing, because the day before it finished without issues, but today I have same issues
I deleted that repo after compiling so we can't see logs of successful build

But we have logs from GH actions from this repo after you created this PR
https://github.com/werman/noise-suppression-for-voice/actions/runs/9017804308

Seems like, yesterday it uses different MSVC version
In this workflow it uses 19.38.33135.0

[pengzhendong]

@DmitriyFrogo

It doesn't build all plugins after I created this PR:

noise-suppression-for-voice/.github/workflows/githubci.yml

Line 15 in c1cf430

# Building all the plugins doesn't add information but adds time,

[DmitriyFrogo]

@pengzhendong
Sorry, I can't help you with that, I'm even not a programmer :D
I couldn't wait when @werman merge this PR, and just forked your repo yesterday, enabled and started release workflow, it builded three files for me: linux-rnnoise.zip, macos-rnnoise.zip, win-rnnoise.zip, I downloaded win-rnnoise.zip, and deleted repo :D
Yesterday it works, today not

[DmitriyFrogo]

I see OBS also have issues with rnnoise compilation with newer compilers and they fixed it
Maybe it's similar issue?

obsproject/obs-studio#10665

[pengzhendong]

I see OBS also have issues with rnnoise compilation with newer compilers and they fixed it我发现 OBS 在使用较新的编译器进行 rnnoise 编译时也存在问题，他们修复了该问题 Maybe it's similar issue? 也许是类似的问题？

obsproject/obs-studio#10665

Not the same issue. The issue is about the LADSPA plugin, not rnnoise.

[werman]

Thanks @pengzhendong, I updated rnnoise myself since I wanted to make the update cleaner: https://github.com/werman/noise-suppression-for-voice/commits/develop/

Downgraded MSVC to 2019 to solve compiler crash. For now BUILD_RTCD is not used on ci since that MSVC doesn't support sse 4.2.

Could someone test the build (https://github.com/werman/noise-suppression-for-voice/actions/runs/9140756650)?

[XCroatoanX]

Could someone test the build (https://github.com/werman/noise-suppression-for-voice/actions/runs/9140756650)?

I can try and test on arch, if needed

[pallaswept]

Thanks @pengzhendong, I updated rnnoise myself since I wanted to make the update cleaner: https://github.com/werman/noise-suppression-for-voice/commits/develop/

Downgraded MSVC to 2019 to solve compiler crash. For now BUILD_RTCD is not used on ci since that MSVC doesn't support sse 4.2.

Could someone test the build (https://github.com/werman/noise-suppression-for-voice/actions/runs/9140756650)?

I don't normally use VST, so I'm not a good test for that (because I would not really know if it failed) but I did test the ladspa version on tumbleweed, and it works great!

Thanks SOOOO much for updating this repo! It's a really big deal for people using system packages, who nearly always get their sources from here rather than a fork.

Also it is nice to know that you are alive and well :D Thanks so much for all the annoying noises we didn't have to listen to because of you! <3

[pallaswept]

For now BUILD_RTCD is not used on ci since that MSVC doesn't support sse 4.2.

FYI I built this as an opensuse package, with BUILD_RTCD, and it works fine.

[werman]

@pallaswept Thanks!