Reverb removal?
Moderators: MattKingUSA, khz
- raboof
- Established Member
- Posts: 1855
- Joined: Tue Apr 08, 2008 11:58 am
- Location: Deventer, NL
- Has thanked: 50 times
- Been thanked: 74 times
- Contact:
Reverb removal?
Like noise removal, reverb removal is of course 'impossible in theory', but there might be options in practice
I have a spoken voice fragment recorded in a badly reverbing room (https://www.youtube.com/watch?v=IgGru4BpQwE). I've tried getting the reverb down with 'simple' techniques (multiband eq, audacity noise removal) but that didn't lead to satisfying results.
It seems like there's some room for specialized DSP tools here, since the reverb definitely has different 'characteristics' compared to the signal. Does anyone know of FLOSS tools in this area?
I have a spoken voice fragment recorded in a badly reverbing room (https://www.youtube.com/watch?v=IgGru4BpQwE). I've tried getting the reverb down with 'simple' techniques (multiband eq, audacity noise removal) but that didn't lead to satisfying results.
It seems like there's some room for specialized DSP tools here, since the reverb definitely has different 'characteristics' compared to the signal. Does anyone know of FLOSS tools in this area?
- milkii
- Established Member
- Posts: 477
- Joined: Tue Jan 05, 2016 9:08 am
- Location: Edinburgh
- Has thanked: 92 times
- Been thanked: 91 times
- Contact:
Re: Reverb removal?
There are some noted on https://gist.github.com/unfa/c0fe13ae9b ... 5d8a86a299
they/them ta / libreav.org / wiki.thingsandstuff.org/Audio and related pages / gh
Re: Reverb removal?
You're never going to get a "natural" enough dry sound without having a "fft analysis" (like what you'd use in a convolution reverb) of the room in which the sound was recorded. My experience is that trying to remove reverb from a recording is more trouble than the final result is worth. I just toss those samples away.
p.s. Audacity's noise removal algo is bad. Adobe's Audition is probably the most useful one I've tried.
p.s. Audacity's noise removal algo is bad. Adobe's Audition is probably the most useful one I've tried.
Author of BackupBand at https://sourceforge.net/projects/backupband/files/
My fans show their support by mentioning my name in their signature.
- raboof
- Established Member
- Posts: 1855
- Joined: Tue Apr 08, 2008 11:58 am
- Location: Deventer, NL
- Has thanked: 50 times
- Been thanked: 74 times
- Contact:
Re: Reverb removal?
Nice, thanks!milkii wrote: ↑Wed Jan 26, 2022 3:35 pm There are some noted on https://gist.github.com/unfa/c0fe13ae9b ... 5d8a86a299
Jotting down some notes:
https://github.com/helianvine/fdndlp - python - used a bunch of CPU and memory but didn't seem to have any effect? (literally: it doesn't just sound the same but the waveforms look identical to the last bit)
https://github.com/sjlee7/speech-dereverberation - python & tensorflow - requires tensorflow.contrib.layers which doesn't seem to be packaged for NixOS AFAICS
https://github.com/fgnt/nara_wpe - python & numpy/tensorflow - did not start for me
https://github.com/rishikksh20/hifigan-denoiser - python & tensorflow - needs pytorch with CUDA, don't have that yet
https://github.com/zehuachenImperial/SkipConvNet - python & CUDA - not tried yet
https://github.com/sas91/jhu-neural-wpe - python & CUDA - not tried yet
https://github.com/HoerTech-gGmbH/openMHA - python & matlab & C++ - not tried yet
https://github.com/giacobello/GroupSpar ... erberation - matlab - not trying, I don't know matlab
(I'll update if I have more progress)
Re: Reverb removal?
"More"? I think you mean "any".raboof wrote:if I have more progress
Author of BackupBand at https://sourceforge.net/projects/backupband/files/
My fans show their support by mentioning my name in their signature.