I would like to request an update/improvement to the Voice Separator stock plugin in Nuendo.
In its current state it feels very outdated and is nearly unusable for professional dialogue work. In many real-world situations (especially complex ambiences, music, room tone, traffic, crowd noise, etc.) it often misclassifies the dialogue as background and fails to reduce background noise cleanly. The result is inconsistent and introduces obvious artifacts.
It would be great to see a modern implementation of Voice Separator.
I’m a proud user of Nuendo for years. I really love working on it, and am also super happy with every major update, but as a DAW focused on audio post, I think it deserves a functional real time denoise solution. I’d buy an update just for this alone.
So that’s all, I really hope you are already working or plan to work on this.
Well of course there are 3rd party solutions. Besides, I’ve been having great results with Spectralayers actually, but it’s not real time. I love the Steinberg ecosystem honestly, and would love to stay in it as much as possible.
I think Steinbergs’ greatest strength is in, the VST format and their cooperation with 3rd party developers, rather than what Steinberg do with the format themselves.
After all, most other plug-in formats are derived from the VST specification in any case so sticking with the DAW that Steinberg makes, only means better integration.
I believe the only issue Steinberg and the 3rd party developers currently have, is with ARA.
Still, I wouldn’t mind knowing what is out there, in terms of Voice Separation plug-ins.
Also, what do you mean by “modernise”, isn’t the plug-in VST3, or are you talking about features?
For voice separation, I have had great success with these: Accentize DXRevive Pro, Acon Digital’s Extract Dialog 2, 4Lo Voxium or Supertone Clear. I would love if Voice Separator by Steinberg worked well but it sounds terrible to me.
Not a specific voice separator but the Unmix Noisy Speech algorithm in Spectralayers is pretty amazing.
Yeah I’m also getting really good results with Unmix Noisy Speech - they improved it dramatically with the last release of Spectralayers - and that is my main tool right now. Voice seperator sounds quite terrible though, and it would be so great to have a solid stock real time processing tool inside Nuendo. That’s this request is about (:
I mean by modernise is, improving the algorithm. Right now it’s very old school. It is kinda working with static backgrounds, but if it’s even a tiny bit complex, dialogue starts to get misanalyzed.
I think this is exactly the type of tool where even though it’s nice to have it as a stock plugin, most professionals at this point already have 3rd party tools for it. DogandPony’s list is spot-on.
Now, because this has been an area of pretty significant development over the past say five years it’s the type of plugin where it’s ok when it’s released, but then surpassed by other plugins a year later, and the improvements are significant. This means, I think, that a lot of users “try it out” for a few months on a few projects, but inevitably abandon the plugin quickly because the alternatives are so much better.
Therefore, I think Steinberg continuing development on this is probably a waste of time. Odds are non-zero that the plugin won’t be cutting edge technology and that it will be surpassed by others again, putting users pretty much back in the same spot, which means it’s a waste of time and resources.
More importantly though, this plugin to me feels like yet another case of introducing something that’s ok and then abandoned instead of taken to completion (or close to it). I mean, how many updates have we seen so far to Nuendo, and how many of this plugin? To me it would make sense that when a new function or plugin is released that it gets updated in following updates to the software until it is at least on par with 3rd party options or the competition. And to me as a post engineer that has to do dialog editing I’d much rather they spend development cycles on speech-to-text functionality instead, since that doesn’t integrate nearly as easily as a 3rd party (denoising) plugin. Alas speech-to-text is another example of a cool new feature with limited use seemingly getting little love.
All of the 4Lo plugins have been impressive, they’re developed by a Hollywood DX editor. Only downside is they put a pretty heavy hit on my Mac Studio M1 Max. And good point, I’d rather see Steinberg put resources elsewhere versus into a plugin like voice separator.
Well I am not looking for perfection or something that would compete with high end 3rd party tools here. Voice seperator is there, and also we know that Steinberg has good algos, in Spectralayers. I also know that a realtime algo is very different with an offline processing algorithm. I’d just be happy if Voice Seperator was working decently and would prefer to use it.