mmm…I don’t see that happening. Why? Several reasons.
Spectralayers as a gui for your own preferred loaded github routines of the past two days? Nah, not SB or Robin’s responsibility to create a “host”…that ultimately a percentage of “characters” would say, "I only wanna load github routines, therefore, I only wanna pay $18 for the gui…stuff like that
Robin already spends months tweaking the recipes that he incorporates into any given version. The results of that are exactly “why” one may decide to buy SL if it’s a better match for a user’s source material.
Other reasons that I don’t have time to opinionate at the moment
I may be misunderstanding your point but IMO DeRustle is feasible and could be helpful. FYI, for example, Acon Digital’s Acoustica features a DeRustle Dialogue plug-in. I’ve never used it in earnest but I’m sure something similar would be worth having in SL.
I’m talking specifically about Todd’s request to have a module that can de-rustle against anyother sounds and I don’t think that is presently possible/practical.
Against dialogue/vocals? Sure, as you said and we all know we have it already.
Sure, in hindsight, we essentially already have what I’m asking for using the workaround of capturing a noiseprint and running DeNoise as @Joey_Kapish has suggested to me for other purposes. We don’t really have an “adaptive” machine learning yet in SLP (nor am I saying we would ever have one); that noiseprint capture is going to remove the corresponding capture from the overall audio users ask SLP to process and ignore unwanteds that aren’t in the capture. In order to retain and re-balance the removed noise to reduce artefacts we’d need to do some reverse phase cancellation to keep whatever audio gets processed out. By then, heck, we’ve probably done what is needed manually (maybe not, long form takes a long time in Post).
Obviously, UNS does an incredible job as Henrique has said; yeah, I can’t imagine dealing with IVs without UNS, cuz it’s incredible. Clearly there are benefits if UNS could give users some control over what audio is wanted human voice and what is unwanted noise (as we’ve already discussed in this thread).
So, maybe what I’m asking for is a Unmix Denoise which would not be destructive and might take a bit less time (not have to carry out the reverse phase workflow). Really, I’m happy to live with the workaround if SLP is better at Unmix Multiple Voices (Speech) and Unmix Instrument and Unmix Chorus. Those modules are already in SLP and IME do not perform well at all.
Fact is, my manual example of DeRustle for the screenshot I shared earlier works so well, that there’s no need to add such a DeRustle function into SLP at this time. And, further, asking for DeRustle just because other applications include such a process is certainly not a valid argument, I apologize for the diversion
Personally, I wouldn’t necessarily adhere to that point of view. The more the merrier is the way I’d look at it!
IMHO that may be true but I’d still suggest that a DeRustle function is perfectly valid as a general request… and there’s absolutely no need to apologise!
Voice Denoise is a destructive process, isn’t it? That’s why I never use it. Looks like I should experiment. I should think the reason I don’t is because I assume UNS doesn’t add any audio to either of the separated audio layers…BUT, I don’t know that for sure, I’d have to carry out time-consuming null tests to varify (hoops I’m unlikely to jump through) .
Say for example using the UNS noise layer mixed with Voice Denoise result of the same audio; I would expect some audio to be completely deleted and some audio to “double”, hence stepping farther away from the original recording and potentially introducing artefacts.
I guess you could view it as destructive but you could always duplicate the main layer before processing, if you were worried about losing wanted data.
Depending on the kind of noise you are trying to remove from the voice and of course the recording in general, Voice DeNoise generally produces outstanding results, largely artifact free. It’s fantastic for taking out wind sounds, trees rustling, river water sounds and so on. You can of course preview what’s being taken out by selecting ‘Reduce Signal’ and clicking on the Preview button, before actually clicking on ‘Apply’ - this will allow verification of how much voice is being lost if any.
I want to point something out, you cannot upload reasonable examples here on this forum! I believe the administrators/moderators STRATEGICALLY implemented this limitation so that issues/problems like this can never be addressed nor confronted. I believe their(administrators/moderators) intentions of doing this is so that whenever a clear problem is demonstrated it can easily be dismissed by the limitations.
Since @venator995 is new to the community, who knows if they are even allowed to post examples at all.
By separation feature I was meaning a general term for unmix sound, unmix choir, unmix crowd noise etc.
I personally would rather upload a cut video to youtube and give a link but again is this allowed? Anyway in that case I would be able to give a link even in PM of you would like so.
No, by “separate vocals” I was intended to mean Unmix multiple voices. So far I had no luck with that one.
Yes, of course you will be able to link to allowed sites like YT. As @JuergenP has said, you just have to wait for your trust levels to elevate on this forum.
That said, mods have locked a thread on me once they deemed I had nothing more worthwhile to add, which I did not agree with, oh well…not a big deal, I think everyone reads with a certain amount “mood” being delivered by posting members and can easily get misunderstood
There was a lot of posts since I last saw this topic, but I would just like to clarify my statements, cause I believe I have been misunderstood:
In no shape or form I believe a deRustle module is not welcome or redundant in SL.
I was merely and solely replying to @ctreitzell regarding a specific de-rustler that could be used on anything as opposed to a “voice derustle” or something similarly specialized. Furthermore, I was specifically saying that I don’t think they are, in that form, viable as of now, not that I dismiss their usefulness or merit.
That’s it. I love derustlers, I think they are useful, I think requests for them are valid, I think Spectralayers should have a DeRustle module and I think no one should be apologizing.
As long as it’s stable, which it currently is not when multiple models are loaded.
Without a serial workflow to load and unload models into GPU memory, users will continue to experience crashes when the loaded models exceed the available memory. For the default Unmix Song process, SL really needs at least 24GB GPU memory to be stable due to concurrently loading multiple large models. The developer uses a 24GB GPU card. The specified min requirement is bullshit if you expect to use Unmix Song via GPU.
