FR: Spectral settings: multi resolution mode

[vectorwarrior]

The problem with spectral effects is the buffer setting is always a trade off, low buffer settings have good transient response but poor low frequency response. High settings have good low frequency response but poor transient response (creating smearing). This affects the quality of all spectral effects.

I would like to suggest a special multi resolution mode that has higher buffer size for low frequencies, with increasingly smaller buffer sizes in the higher frequencies. This way the lower frequencies can get the better bass response and the higher frequencies will get the better transient response. The best of both worlds.

Thoughts? I think I would be great for spectral dynamics, morph, transformer, etc.

Yes, you can attempt to do this on mxxx with crossovers but ideally you want the frequency ranges aligned properly and for the latency to be accounted for, both of which are difficult to do manually in mxxx (I have tried). I'm sure Vojtech could figure out the optimal way to set it up.

Isotope Rx can do something similar, I can investigate if people are curious.

[Alphacodex]

Sounds like a very welcome feature for the spectral effects in the melda arsenal. I use spectral dynamics on almost every mix I do, so every optimization here is welcome!!

Cheers,
Codex

[MeldaProduction]

Actually that is what super-resolution in analyzers does, and it is VERY VERY complex and takes a LOT and LOT of CPU. I don't even know if that could be adapted to resynthesis (processing, not just analysis). It would be cool, but I'm afraid there's just not enough time now...

[vectorwarrior]

Interesting! Ah that sounds really cool. I hope you can find the time at some point because it would boost the quality of all your spectral effects! CPU is rarely an issue for me, and with freezing it's a non issue, so bring it on

[MeldaProduction]

It's hard to say if it would even work. It is ok for analysis, but for resynthesis there just could be extremely big problems... Basically I think this way it would disrupt the coinciding phases between multiple resolution blocks and that would just sound horrible . Of course there's a way to literally analyse individual frequencies, like super-high resolution vocoder, I was always thinking about it, the trouble is, even your computer (maybe not even a supercomputer) wouldn't handle that .

[vectorwarrior]

Wow, you shouldn't have said that as now I'm really excited by it! Even if it was 'render only', like the sample rate and interpolation settings, that would be a big deal for me. I'm using morph a LOT for my current game project, so would be very very interested in this.

Anyway, I'll leave it alone now, I know you're busy with v11 and TurboReverb, and if it's worth it I know you'll investigate. Thanks for the consideration!

[MeldaProduction]

Hehe don't be excited, the chances are slim imho .

[vectorwarrior]

Just going to bump this up. The main reason being that Zynaptiq have just launched Wormhole and I've been trying it out compared to Melda's MTransformer and MHarmonizer. Wormhole's pitch shifting sounds pretty tight and clean compared to melda's... so I did a test in MXXX with multiband transformers or harmonizers using smaller buffer settings for the higher bands, and you can get an improved sound with this approach, but it's not as clean and tight as Zynaptiq's. So, I'm going to guess if this feature-request could be added it would drastically improve the quality of all Melda's spectral effects; Spectral Dynamics, Morph, Transformer, Harmonizer, the new spectral crossovers, etc.

As CPU is a concern, I think an 'offline only' option would be fine as it shouldn't effect mixing decisions too much, it'll just sound better. But with computer power these days it might be okay at realtime.

[MeldaProduction]

Well, to be honest, I simply have 0 time these days, so no chance in the near future... Anyways I don't think this is "the way" really, but who knows, I'll think about it.