Hi all=)
Here are two recurring annoyances that should be addressed by Steiny in my book…
-
Variaudio note detection
-
Hitpoints detection
-
Here is an example of how Variaudio handles a very simple 1-voice bassline.
- In other cases, variaudio simply ignores huge portions/chunks of audio that are clearly identifiable. Variaudio should not
simply “ignore” what it cannot “read” and create a “generic chunk” that is moveable/editable (even if it is not
the correct note)
The “segmentation” tool also cannot always “cut” the material… (it simply ignores the cut)
-pic below-> simple 1-voice bass variaudio detection / dry render
-pic below-> simple 1-voice bass variaudio detection / render with channel settings (vanilla channel)
-pic below-> simple 1-voice bass melodyne detection (closest to original)
- Hitpoints are always wrong because the threshold of detection is too high…
- Starting points (first hitpoint) of recorded audio is always too late. The first hitpoint of a recorded soundbite is
never on point, but always starts a good portion into the sound file. When you zoom in to the waveform, you
can see this. (maybe this is why transient rich material like kicks always sound weird when stretched to tempo
because it “cuts off” the first part of the sound due to a faulty hitpoint detection)
-pic below-> First hitpoint detection - zoomed waveform / how it (wrongly) detects every time
-pic below-> First hitpoint detection - zoomed waveform / manually corrected-how detection shold be everytime.
