Izotope RX11 Dialog Isolate CPU Usage // Tech Support's Response

[@run_animal_run]

Hey there! I also experience that my macbook m2 max is working heavy while rendering dialogue isolate or music rebalance. But I also hear a huge improvement of algorithms, so for me this works out. I wanted to suggest something for your workflow. How about having the untouched audio clips above the ones that you processed? so if you run into any problems you could switch out the processed part with a unprocessed part or process that paticular part separately? Was just thinking of that...

[@jesse.mccune]

I haven't used the older music rebalance to compare, but the new Dialog Isolate is noticeably better than it was in previous versions, aside from the performance issue. I like your idea, but I'm happy to be able to work in a completely non-destructive manner. For me and the work that I do, efficiency is more important. I have moved away from using the RX editor in favor of only using plugins for the last 18 months. The two things I have learned from this experiment is that I can accomplish everything I need with plugins and not making that trip to the editor saves me a decent amount of time.

[@Tomaslav16]

Уважаю тебя за твою точность и въедливость в детали и тонкости технических процессов!!! ты пожалуй единственный в медиа пространстве кто так по мужски точен!!! Продолжай обязательно! Твоя информация и нам "Экономит кучу времени" при выборе инструментов работы со звуком)))

[@jesse.mccune]

Thank you for the kind words and I'm happy that you're finding my content helpful.

[@ChrisPFuchs]

Neural Network Denoisers are generally used as Audiosuite or on the Dialogue Bus for Post Audio. It's just too intensive to run on every Dialogue Track. Podcast Editing is perhaps a bit unique in that you need far less Dialogue Tracks so could expect to get away with it. I think the Real Time DX Isolate serves its purpose fine for Post Audio and does sound quite good, but the render time for the 'Best' is annoyingly long and I can see why not being able to run multiple instance of it is a big negative for Podcast. I personally offline render my heavy noise reduction; the two best Neural Network Denoisers in my opinion run off of dedicated AI accelerators and are only audiosuite or offline rendering at the moment.

[@jesse.mccune]

What industry do you work in? Podcast editing is definitely a unique situation when it comes to dialog editing, for many reasons. I can see why running tools like this as Audio Suite on buses would be the norm in many situations. What makes podcast editing different is in most cases, we're dealing with low paying clients with short turnaround in an industry facing a lot of downward pressure and commodification of our skillset. Between AI and sites like Fiverr and Upwork, clients are increasingly asking for more while spending less making efficiency the number one goal to remaining profitable. I'd approach things differently if I were working on a project that had higher standards than podcasting where good enough and affordable is all that matters. What are the two best Neural Network Denoisers you are referring to?

[@ChrisPFuchs]

@@jesse.mccune Hush Pro is generally regarded as the best Dialogue Denoiser in the industry at the moment and runs off of the Apple AI chips on their gpu. According to the creator, this allows orders of magnitude larger model sizes to be ran as opposed to cpu based Neural Denoisers and I think it shows. Audio companies like Izotope have to develope for a much wider range of systems however, as AI chips are really only found on a few Apple devices and Nvidia GPU's. This means running it off of the CPU. Auphonic is the other one; the 'Dynamic' Denoiser runs off of specialized hardware in the cloud. I know you're a little opposed to cloud based rendering, but the quality is good. It also has extremely quick render times and you can actually have multiple files being processed at the same time. I do Podcast editing regularily for a company, but also do other post audio work.

[@jesse.mccune]

@ChrisPFuchs Thanks for the response. I like Hush, but since I'm not a PT user, I don't have access to the Pro version. The biggest issue with the regular Hush is that there's no preview function or a way to render a small section to test the settings. It becomes more of an all-or-nothing endeavor where I guess on a setting, render and listen back. I've had a couple conversations with Ian and suggested this to him and he seems to think he can bring Hush into VST/AU format, but needs to figure out a couple things to allow for that. When you say you do podcast editing for a company, are you a contractor for an agency or were you hired by a company to edit their podcast?

[@ChrisPFuchs]

@@jesse.mccune Oh yeah, I can see how that'd be a deal breaker! Hopefully the realtime vst version is good. Ian seems like a cool dude. But to summarize my original comment about RX 11, it just seems like it was engineered within the constraints for Post Audio; a solid realtime denoiser that sits on the DX bus and a slightly 'better' offline algorithm that works well for Audiosuite processing. I think it's 'fine' in that sense, but as you point out, its weaknesses definitely show a little for Podcast when you have long files to render or tryimg to out use it on multiple dx tracks.

[@Khaled_Omar1994]

great work and very important issue

[@jesse.mccune]

Thanks. I know it may not be an issue for many people, but for someone who prefers to work directly in my DAW, this is a deal breaker until they better optimize the plugins.

[@Khaled_Omar1994]

@@jesse.mccune I really appreciate your effort and your wonderful content, continue to the top😍😍

[@frjr7402]

When I open RX Repair assistant it causes horrible lag and distortion in reaper 7.30. I’ve tried multiple troubleshooting methods and nothing. (My computer is above the min requirements for rx 11 elements)

[@jesse.mccune]

What you’re describing sounds like there aren’t enough resources to process the plugin. I’m not familiar with Reaper, but there should be a place in Settings to change the buffer size. Try increasing that to ease the strain on your CPU.

[@omerylmaz958]

hahaha I love it adobe voice destroyer

[@jesse.mccune]

I can find uses for most noise and reverb reduction tools, but that one changes the voice too much if there's too much noise. It makes them sound they sucked on some helium. It doesn't even know how to handle decent to good recordings. Case in point, I tested a recording my sister made that didn't have any noticeable noise and needed a little reverb reduction. The audio Adobe returned wasn't even recognizable as her. With these tools, the voice should still sound like the person after being processed, not an AI caricature of the person, so Voice Destroyer seemed like a good reference.

[@omerylmaz958]

​@@jesse.mccune to be honest, this feature used to be better when it was first launched, but seeings things like these everything is changing to a post editor kind of thing. The human aspect will never go away. Its just the person with more skills will be ahead always

[@jesse.mccune]

@omerylmaz958 Agreed. I haven't tried it out since they moved it into their subscription package, but I noticed that each new algorithm seemed to be worse than the previous one. That was always a head scratcher to me. It leaves me wondering what they are training it on and who is testing it if they think the results are getting better.

[@synthetic144]

using the wrong tool for the job

[@jesse.mccune]

Could you elaborate? Why is it the wrong tool and what is the right tool for the job?

[@synthetic144]

@@jesse.mccune dialogue isolate is for extenely noisy environments, for most jobs I use spectral denoise for constant noise i can profile otherwise dialogue denoise for more noise suppression type jobs, also dehum, so depends on the job but i wont go for dialogue isolate for general type suppression and certainly wont slap it on a track and call it a day

[@jesse.mccune]

Before RX11, I would agree with you. I'm curious what type of work you do. For me as a podcast editor, I'm looking for efficiency and the new Dialog Isolate appears to be Izotope's response to tools like dxRevive Pro and Supertone Clear. They even promote it as an all-in-one tool for removing noise and reverb from dialog. This sounds like it's meant to be the right tool for my purposes. My clients don't pay enough to go into the RX editor to manually dial in Spectral De-noise and any other work that would need to be done in the RX editor. There are other jobs where the client is paying a premium for that level of work, but it's not common in the podcasting space.