Stereo Tool

Bojcha: I may have an idea to detect when voice vibrations occur. Will have to check if I'm correct.

Basically, what I think voice vibrations are is this:
- A voice consists of many tones of different volume
- If the bass is loud, some (the loudest?) tones are reduced more than others during peaks in the bass level.
- This differs from normal volume vibrations (say if you play 2 sine waves), because - I'm guessing here - the human ear is probably much more sensitive to tonal differences than to overall volume differences.

This would mean that while normal sine waves (bass + high, or bass + high frequency noise) would still sound good, bass + certain combinations of sine waves might get problematic.

If I'm right the solution would be simple: Constantly (when needed) clip the highs at a lower level to reduce tonal differences (basically the vibration gets constant, which does cause a reduced quality in the voice). I actually did something like that in the old trumpet filter (which also worked on female voices) in version 4.22, but I've removed it in version 5.00 after introducing the phase rotation filter. In that version I did it too much, and I think I might have thought of a reliable way to detect when vibrations occur.

Exactly.
I did't know that "highs" may affect voice (at for example 500Hz) to vibrate because of Lows !(?)

When i hear voice to vibrate ("clip at" is at 90%) i manualy put to 60% and no vibration, or,
I lower that band wjere is voice in multiband for mostly 350Hz to 1.4kHz and again no vibration.
But ofcourse it's better lows to be lowered.

Please check PM.

Actually, voice is at all frequencies starting around 300-500 Hz or so, going upto over 10 kHz (but most is below 4000 or so I guess).

If the freqs around 3000 are lowered regularly, and the rest stays constant, the vocal content changes. Evolutionary, it can imagine that it's far more important to recognize such differences (which determine for example who says something, or with which emotions), than the total volume (which can differ due to many not very interesting conditions). So the human ears will probably recognize such differences much more easily.

Fix for the bug I just mentioned (some audio was incorrectly filtered out when the new slider settings were changed, and there was also something wrong with audio around 500-900 Hz at the default settings).

Clean/Punch is now re-enabled for bass frequencies only.

Winamp DSP plugin: http://www.stereotool.com/download/dsp_ ... 510-38.exe
Stand alone version: http://www.stereotool.com/download/ster ... 510-38.exe
VST version: http://www.stereotool.com/download/vst_ ... 510-38.dll
VST version (No SSE2): http://www.stereotool.com/download/vst_ ... 510-38.dll
Command line version: http://www.stereotool.com/download/ster ... 510-38.exe
Linux command line version: http://www.stereotool.com/download/ster ... BETA510-38 (not built)
Linux GUI version: http://www.stereotool.com/download/ster ... BETA510-38 (not built)

I've been checking this really difficult song that Bojcha posted a few days ago, and I've now discovered that the voice vibrations are occurring at far lower frequencies than I though.

With the fixes in BETA38 I seem to be getting better results with the following settings:

- Clip at 90% (or lower)
- From 0 Hz upto: 150 Hz
- Then rize to 100% at: 250 Hz
- Highs vs. lows protection runs from: 325 Hz
- to: 400 Hz

Hm, try this:
- Make sure Phase Rotation is enabled (that should ALWAYS be enabled when Loudness is used)
- Set 'start freq' to 120 Hz!
- Set 'initial speed' to 400 Hz.

Phase Rotation currently starts at 700 Hz, but that's too high since (as Bojcha has just reported in his PM) vibrations already start at lower freqs.

nope..
This not helping... actually any Phase Rotation setting does not help.
In my preset I already keep PhaseRotation start freq at around 250Hz.. and initial Speed at 1000

"Clip AT" need to be 90% and to go down in that very special moment almost to 50%!

btw i re-uploaded this track ripped from CD at 320kbps
http://www10.zippyshare.com/v/85309697/file.html

Which moment is that? I've checked some waveforms and I think I have a very good idea of what needs to be changed, but I want to check it for the most difficult sound to see if I'm correct. If I am, a smarter version of the old Trumpet filter should do the trick, and I wouldn't even have to reduce the bass level.


2 new remarks:
- Setting Highs vs Lows protection to 1000-2000 Hz gives a much better protection of the bass against loud 'S' sounds, however it allows more voice vibrations. I'll need to find a solution for this.
- I think I need to restore the filter that I removed that boosts bass when loud S's are present. It might cause some distortion, but it seems to be needed to avoid bass level drops during loud 'S' sounds.

Moment when voice is distorted by bass.

I mean there are a lot of those, and most are solved by enabling Phase Rotation or reducing the clip level just a bit. I assumed you had a time in mind for 'that very special moment'?
For what I checked so far my idea should work. (I've checked the U2 track 'with or without you' with lots of bass and a beep tone (also in the Orban samples), and it sounds reasonable with clipping at 100%, and very good at 80%.)

The voices that I checked so far have huge repetitive spikes - even after Phase Rotation - which explains why they sound bad when bass is added. That could be easily fixed by just cutting off those spikes altogether.