Hi all,
Many of you requested audio de-bleeding capabilities for the next version of SpectraLayers.
If you have such examples I’d be interested in crash-testing them against potential algorithms.
You can send me links to your samples (including 2 tracks or more overlapping) here, or in PM, or to contact [at] divideframe.com (but no file attachment over 5MB please, in that case prefer online file hosting).
Thanks !
Hi Robin,
Thanks for taking up this request. That is really great! I will be more than happy to provide you with examples of what I am looking for. I will need to send the files to you through WeTransfer. Can I use this e-mail address for that?
Please let me know, and I will send today,
Magnus
Hello Robin,
Awesome! thanks for tagging me as indeed i’m very interested in this feature.
I just emailed you, let me know if you received the files.
Cheers,
I’ve sent you a message through the forum, Robin.
Best,
Magnus
Isn’t this essentially what the unmix stems would do? I’ve got the Essential version and I have been successful at de-bleeding vocal tracks rather easily. And I’ve done it to remove vocal bleed from instrument mics too making it possible to even tune vocals from live recordings.
You can use that feature to de-bleed vocals indeed. But bleeding is not limited to vocals, one common occurrence is recording drum kits with one mic per instrument for instance.
I’m only commenting here because I (who has done some coding in the past) feel like a “de-bleeding” feature is completely unnecessary. The “imprinting” feature within spectralayers feature is pretty much the same idea of “de-bleeding”, and I dont feel it is necessary to waste precious resources on implementing features that wouldn’t really benefit the end user. If anything, I would highly recommend improving the “imprinting”(casting/molding) features rather than waste time implementing something that is not necessary. I feel that “imprinting” is something that can be improved significantly and maybe perhaps something that can be done in real-time. I would not recommend any “de-bleeding” feature because when you use the “unmixed stems” feature and “imprint” the stems, you get results very similar to “de-bleed” (meaning, it’s the same logic).
Hi Joey,
Wow, how is this helpful? Many people are requesting a debleed function, and Robin is willing to take us up on this request. So why would you want to throw a spanner in the works here?
For people who - like me - do a lot of live (studio, in my case) recording, the option to debleed for instance the drum kit mics from say guitar, bass, and other instruments, would be a dream come true. This does not work at all well today, with the current Unmix functions in SL Pro, not even for drums. So if Robin offers to take the AI through a training session for this, that would be extremely welcome.
What may seem “completely unnecessary” to you, may be absolutely essential to somebody else. Right now, SL Pro is very much geared at remixing. Debleed options would be more than helpful when mixing.
/Magnus
Because (like I said and mentioned) “imprinting” is basically the same thing and same idea as “de-bleeding”. I feel strongly about not adding a “de-bleeding” feature because development time is precious and to waste development time doesn’t help the end user. To implement a feature that already exists doesn’t help you nor me and it hurts us(the end user) because the developer is wasting time doing something that doesn’t benefit us. The “imprinting” feature is actually more powerful and useful than “de-bleeding” (when you think about it)… The wise thing to do (so we can all benefit and all win) is to “improve the imprinting feature” (maybe perhaps real-time imprinting) not waste time on stuff that is not needed… Also (from my experience) when I did use the “de-bleeding” feature from other programs like Izotope RX, it made the end result of the audio sound undesirable thin and left me disappointed. Same thing with audionamix’s “de-bleed” feature, yes it did remove some of the artifacts but the end result sounded very thin and it’s very noticeable when you remix. I dont believe “de-bleeding” is the correct approach to a cleaner stem and because “spectralayers pro” is unique as is shouldn’t head in the direction of what other programs are doing.
Although there are similarities with the imprinting process, the imprinting process is quite destructive and aims more to create space in frequencies (mastering or creative use).
A specific de-bleeding process is still needed to properly evaluate the amount of frequencies that needs to be removed from one layer to another without affecting too much the main content of the target layer.
Also different mics may have different frequency response curves, which also needs to be evaluated. A de-bleeding process can also account for room reverberation if needed. So the two (imprint and de-bleed) have reasons to exists.
Okay, the only thing I’m worried about is development time though. However, I will mention this, I feel “de-bleeding” is also destructive as well (that is why the audio sounds incredibly thin), many audio engineers/remixers have also mentioned the same thing, “de-bleeding” is not a pleasant result for me. If you can somehow manage to reduce the gain from the other layers while maintaining the core partials/harmonics/overtones then I would be happy for that feature. I would only recommend a “de-bleed” feature if you manage to keep the partials/harmonics/overtones intact without the results sounding too thin (but then it would mean that there is improvising going on, like the idea of “unmix components” where the hidden NOISE is reconstructed, which also means you’re also destroying the phase… So that wouldn’t make sense either).
I’ll send links for examples to you.
“(but no file attachment over 5MB please, in that case prefer online file hosting).” David L
What resolution and track length would be better for testing/training?
How many tracks would you prefer to receive? 2, 3 or more is better?
In my case I have a session where tracks suffer from the bleed from the headphones that a musician being recorded, uses for referencing his part improvisation (containing full tune), despite he was using a Sennheiser HD-280 Pro (suppress -32 dB at 1 KHz, one of the highest isolation specs in its range)… and is quite ubiquitous.
The bleed is quite low volume, if I want to use the parts in some other context/effect it could became a problem, more so in these times, everyday expecting more definition and lower noise.
If this could easily happen where a violin is close miked. I imagine there should be plenty of other solo instruments or usage cases where small recording rooms or small studios would need such de-bleeding capacity.
Very much welcome.
I’m looking for typical examples. So whatever resolution you work with. If you can reduce track length to 30 sec that will save space, but it’s fine if you send full length tracks as well.
Number of tracks: 2 to 4, the most relevants for de-bleeding in your opinion.
Thanks for taking the request! I haven´t specifically asked for this feature, but I have been involved in post asking for this. I agree with those who have brought this feature up - it would be a great addition! Sorry if I can´t contribute with any files though.
Best wishes