It would be nice to have a module that removes wind rumble from recordings.
1 Like
I use “unmix noisy speech” for Wind and microphone rustling
P.S.: It’s also for a large part of the ambient noise that might leak in. You can adjust the level because it’s divided into 2 layers.
1 Like
I also use unmix noisy speech for this. I usually want the freqs above 400Hz. It certainly depends upon which freq component users want to remove/ reduce. 200Hz and below usually have to be removed/ reduced.
That said, it is still a manual process and I typically use transfer tool to transfer unwanteds to a muted trash layer after the unmixing.
@Damian_Kearns clearly you are looking for a faster solution for long form 
I do this too but what if there’s no speech?
1 Like
For music I use Unmix Soundtrack…
Great module!
1 Like
Really great module. My thing is when I need to remove wind from ambiences or car passes or animal noises, there’s nothing I can really UnMix.
1 Like
From looking at the flawed (yet amazing) results we get when unmixing Dialog, which is something easily definable in a ML context and strongly differentiated against ambiences, I guess having something that would reliably discern wind noise from the myriad of other possible noises and timbres in something so widely varied such as “ambiences” is a decade down the line.
On top of that, in my experience removing wind noise is something that usually requires subjective judgement on what parts should be kept in order to achieve a natural result, unles you just replace the audio in the part with another segment without wind (but that’s just plain old editing then isn’t it?). I think that’s cause it usuallly wind rams the mic capsule and effectively ruins the rest of the signal where it’s present. Like there is no other information to bring up alongside it. It’s wind or a hole.
For that reason, I find that currently the best method is to just spectrally edit it out, using Clone Tool, Heal and Eraser (sometimes just reducing the gain is enough to bring it into the background). Its quick enough.
1 Like
My experience is in-line with what Henrique says:-)
I’m working on this right now: on location, outside…distant traffic noise, some clothes rustle (creaking of leather somewhat)…and wind noise
In this particular scene, I want the ambience of the outdoors; therefore, the only component of the wind I don’t want is 200Hz and below speaker breaker and low end rumble; all freqs above I want
I just reduce the low end; but not all of it: transfer tool set to 45%/80% hardness with a fairly wide brush and transfer to the muted unwanted noise layer.
For traffic noise which is living around 200Hz-1.5KHz I transfer a good portion of that to another layer and attenuate…maybe with some envelope editing
The rustle in my case is essentially a series of transients over time which I use magic wand to send 100% to unwanted noise layer and cover with clone stamp if artefacts stand out.
The RX De-Wind Module at least can pull us in the right direction when editing but it’s more or less stupid, as in not much machine learning to it, compared to the Unmix modules in SL. I think this it the kind of thing– what I’m asking for- that doesn’t need to recognize speech, just the wind noise. This way, a future module like this could be used on anything, even things that don’t have speech. As Henrique mentioned, it’s a fairly fraught editorial process to subdue wind. One module likely wouldn’t work well on everything.
1 Like