Spectral Layers 10 Pro Blowing My Mind

[Todd R]

Um, I am blown away at how amazing Spectral Layers 10 Pro is.

Removal of crew chatter on the shotgun mics and retaining the ambient "tone" of the locations recorded is possible

I look forward to using it on music

there's a bit of a learning curve, but, WOW! amazing results

[UpTilDawn]

Pretty remarkable results at separating parts in song, drum and multiple voice extractions so far.
Of course, now I need even more precision and ability to distinguish more instruments (winds and brass, strings, etc. - I'll never be completely satisfied... ).

It's allowed me to salvage quite a few recordings in just the past few months. And I barely use its features other than "unmix"...

[Todd R]

Pretty remarkable results at separating parts in song, drum and multiple voice extractions so far.
Of course, now I need even more precision and ability to distinguish more instruments (winds and brass, strings, etc. - I'll never be completely satisfied... ).

It's allowed me to salvage quite a few recordings in just the past few months. And I barely use its features other than "unmix"...

I'm manually removing crew chatter and retaining ambiance ...shocking how great it is at that

I haven't tried any music until just now...clearly the AI needs to learn about the entire symphony rather than pop song parts!

It's surely coming soon

we've been generating AI text reads for an online course these past few months (not SL)
and the improvements from the summer to the fall 2023 went from unusable to indistinguishable
frankly the AI is better than getting a human reading for difficult text...the AI doesn't flub
tongue twisters like a human...and saves time at the cost of some expression

I mean, I doubt anyone could blind test the difference between human and AI

[Microstudio]

What is that compared to Izotop RX 10?

[UpTilDawn]

They both have their strong points for me.
Where separating parts is concerned, Spectra has the ability to potentially eek out more useful parts and offers more options - like separating drum sounds into kick, snare and cymbals - which it does quite well, depending on the style and complexity. Although I haven't had a lot of need to dig deep yet, it also is quite good at separating multiple voices and lets you choose how many to limit the search to.

Sometimes, so far, RX is better overall and sometimes Spectra is better. At first I thought I might abandon RX for this job. but it quickly became apparent that for my needs, I'd better plan on using both.

[Todd R]

What is that compared to Izotop RX 10?

RX10 Advanced costs $1000.00 more than Spectral Layers! I'm not an Izotope fan, personally...

maybe watch a YT shootout?

[Todd R]

I'm using it a lot more like this guy...my audio is on location on a no budget film...his is "in the studio"

[Todd R]

hmmmm...as far as unmix of music, it's not really there yet for my purposes beyond some amount of deconstruction to then re-record...it's alright at separating drums and vocals...but so much of my homespun music is loaded with layered cross-overs where the AI in something like Spectral Layers needs to be taught which specific instrument is what timbre...

[Todd R]

Nevermind this awful Steinberg video (the audio is terrible in these and the self-aggrandizing nature of some of the commentary is utterly nauseating), the advice in this video's first comment is priceless (careful, the author wants money if he finds his advice regurgitated elsewhere so I shant quote him beyond my quick summary)
This guys says to unmix many, many, MANY times then comp as makes sense. Seems like very solid advice. I am "hacking with manual tools", so I'll try out the multiple passes of unmix to see if that improves things.

[Todd R]

Well, just to say I tried multiple passes of mixing and I didn't experience much joy. I'm certainly getting much further hacking manually. The material I'm working on is a scene where we are mid-interview with the main contributor at the front of his house and a friend of his drives up and the friends dog goes off barking and the crew chatters ...I'm trying to break it down into separate parts...unmix noisy speech works pretty good the first pass...and then does hardly anything (if anything) on subsequent passes of unmixing the noise

maybe working with music yields better results