FabFilter

One idea for potential FabFilter Pro-L3 is spectral limiting, which might use the same FFT approach as FabFilter Pro-Q4's spectral dynamics but with psychoacoustic model (derived from lossy codecs like MP3 and AAC) to shape the gain reduction per frequency (e.g. masked parts received less gain reduction or something else)

Another idea for Pro-L3 is the priority EQ section, which could behave superficially similar to Pro-C3's sidechain EQ section, which depending whether spectral limiting is enabled or disabled, either changes how the spectral limiting process prioritizes certain frequencies (boosting these will limit these frequencies more aggressively and cutting these will "protect" certain frequencies at the cost of more reduction on other frequencies) or in case of spectral limiting being disabled, the distribution of faster transient and slower release stage activities per-frequency (boosting these frequencies will make the slower stage more active on these frequencies, or cutting these will make the faster stage takes over more on some frequencies)

BTW, the Pro-L3's spectral limiting idea is not complete without "Spectral Density" settings from Pro-Q4's spectral dynamics that adjust number of scalefactor bands used for spectral limiting, as well as gain reduction spectrum display on the priority EQ section of Pro-L3 that is absent in sidechain EQ section of Pro-C3 (even on "TTM" style that is actually multiband)

Alternatively, the priority EQ could work similarly to Waves L3-16 multimaximizer plugin when spectral limiting is enabled (to make it more consistent with priority EQ on broadband limiting mode/spectral limiting disabled); boosting a frequency will "protect" certain frequencies at the cost of slightly more pumping and cutting a frequency will allow certain frequencies to be attenuated more with less pumping, especially when psychoacoustic model is either more sophisticated than or differs from ones used by lossy codecs (e.g. MP3, AAC, Ogg Vorbis and Opus) right?

On that note, I've somewhat implemented the idea of priority EQ (for FabFilter Pro-L 3) on my own multistage limiter effect (as a sidechain prefiltering for the first stage, which uses K-weighting though priority EQ in Pro-L3 could be anything):
codepen.io/TF3RDL/pen/emBzNpm

BTW for FabFilter, feel free to omit the spectral limiting in favor of something else (e.g. more intelligent channel linking using an iterative numerical solver algorithm similar to DMG Limitless that works out gain reduction across multiple discrete channels on both stereo and especially surround formats) if you don't need it

I hope Fabfilter include all the things Tone Projects have in their new Uni-L Master Limiter. What it is able to do is extraordinary. Tone Projects have now taken the crown from Fabfilter as far as limiters go.

However, I think a multi-resolution FFT approach (e.g. approximately 4800 samples for lower frequencies, approximately 2400 samples for mid frequencies, 1200 samples for higher frequencies and even 600 samples on even higher frequencies) are more conducive than a conventional FFT approach (as in MeldaProduction's MSpectralDynamics and even FabFilter Pro-Q4 with spectral dynamics on it) for a spectral limiting within probably Pro-L3 limiter, especially with time-frequency resolution tradeoff thing and the spectral limiting in Pro-L3 would probably behave completely different from spectral dynamics in Pro-Q4 as it would probably uses psychoacoustic model to do spectral limiting transparently as I said before

And BTW, since I have been aware of multi-rate constant-Q transform/variable-Q transform implementations like Gaborator exists out there and so, FabFilter could adopt this multi-rate method to upsample only the higher frequencies (e.g. native sample rate/no oversampling at lower frequencies, then lower oversampling amounts in middle frequencies, higher oversampling amounts on higher frequencies and so on) when spectral limiting is enabled while oversampling is also engaged at the same time, rather than downsampling the lower frequencies as would for other multi-rate audio processing

And also, what about multiple spectral styles? Something like this:
* Broadband: Self-explanatory; a good old single-band limiting
* Clarity: Attenuating masked parts less than maskers, which is the opposite of lossy codec models (that discards masked parts) and so, it is basically a downwards version of apulSoft's apUnmask plugin as instead of boosting the masked parts, it attenuates the masker parts, especially under heavy limiting
* Versatile: Using intelligent psychoacoustic model to combine the best of single-band and multiband limiting
* Loud: Prioritizes maximizing loudness while being more perceptually transparent than multiband compressor effects from broadcast processors

BTW, spectral limiting should be turned off by default (aka. "spectral style" parameter is set to "Broadband" mode by default) as usual if added into FabFilter Pro-L 3

And on unrelated note, how about adding "Show individual stages" option for gain reduction metering (gain reduction meter bars and graphs, and probably even gain reduction spectrum in priority EQ section if added into Pro-L3)?

Especially when FL Studio's Emphasis limiter effect shows individual gain reduction stages (first stage and combined first/second) on its gain reduction meter and my own web-based multistage limiter project even has individual gain reduction stages (first stage, second stage and combined first/second stages displayed by default) for both gain reduction meter (on peak meter visualization mode) and gain reduction graph (on graph over time visualization mode)

Though FabFilter Pro-L3 spectral limiting isn't complete without separate spectral shaping amount parameters for transient and release stages, which determines amount of spectral limiting vs. broadband limiting in terms of behavior (e.g. 0% means broadband limiting, 100% means spectral limiting and 50% is somewhere in-between) for each stage to control the tradeoff between maximum loudness and minimal pumping and distortion vs. frequency response preservation

😂😂this dude is still adding features, give him a few days and he will send the code

And what about the spectrogram display (aka. time-frequency visualization) for the graph over time? Especially with gain reduction graphs especially when spectral limiting is enabled (which is basically like the gain reduction spectrogram as in sonible smart:comp 3's spectral compression but combined into one graph), though it could also apply to input and output graphs too for spectrogram visualization

BTW, the spectral limiting idea for FabFilter Pro-L 3 limiter plugin is not complete without "Resolution" parameter, which controls the tradeoff between bass frequency resolution vs. latency and transient smearing (but much less significant in higher frequencies as it would use multi-resolution FFT to improve temporal resolution for higher frequencies), much like linear-phase/spectral dynamics resolution/latency settings in FabFilter Pro-Q 4, probably with "Low" setting have maximum resolution of 1200 samples, "Medium" setting have max resolution of 2400 samples and "High" setting have 4800 samples of max resolution

BTW, the spectral limiting for FabFilter Pro-L3 would make the external sidechain feature of this limiter plugin more useful (e.g. to unmask something during mixing, especially with kick and bass) outside stem mastering cases

And also, I think the bass boost communities would benefit from a spectral limiter (just like multiband limiter and multiband compressor), especially with extreme bass boosted versions of songs on YouTube with some of these either have the rest of the frequencies attenuated by static amount, have pumping artifacts (when using a single-band limiter) or a clipping distortion (when not using a limiter effect, which lets the peaks exceed 0dBFS, which caused distortion when converted to fixed-point audio formats), and that spectral limiting (much like multiband limiting and multiband compression) would allow you to do an extreme bass boost with little to no pumping or distortion as possible at the expense of probably a less amount of bass boost effect as it shapes the frequency response (which attenuated the bass in case of bass boosted songs)

The bot strikes back lol 😂😂 but yes I want Pro L3 so I support the bot