You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
While working on integrating the new preset parser and rewriting the preset rendering process, I noticed some presets were behaving highly erratic or were completely broken (e.g. "shifter - robotopia"). I've tracked down these issues to the beat values being way too high.
The Milkdrop preset authoring guide isn't very specific about the actual value range (>0), but says that the values are expected to be in the range from 0.7 to 1.3 in general and around 1.0 if the levels don't change much.
Looking at Milkdrop's audio analyzing code, these values aren't absolute sums/averages of the current frequency spectrum intensity, but rather slighly or more smoothed relative values to the intensities of previous frames.
@ErikPrantare - you've refactored the beat detection code a while ago. Would you mind looking into this again and implement it in a more Milkdrop'esque way to keep the value ranges in the expected range?
The text was updated successfully, but these errors were encountered:
Spectrum analyzer data is generally very different from what Milkdrop provides. Most presets using custom waves in spectrum mode will render garbage, as the values/peaks are way too high. See many presets in "Waveform/Spectrum" in the Cream of the Crop collection as examples.
Oscilloscope data also has issues when drawing custom waveforms with more than 480 data points (512 is max, 576 samples should be provided for smoothing), there's a small spike at the front which shouldn't be there. Example:
While working on integrating the new preset parser and rewriting the preset rendering process, I noticed some presets were behaving highly erratic or were completely broken (e.g. "shifter - robotopia"). I've tracked down these issues to the beat values being way too high.
The Milkdrop preset authoring guide isn't very specific about the actual value range (>0), but says that the values are expected to be in the range from 0.7 to 1.3 in general and around 1.0 if the levels don't change much.
Looking at Milkdrop's audio analyzing code, these values aren't absolute sums/averages of the current frequency spectrum intensity, but rather slighly or more smoothed relative values to the intensities of previous frames.
@ErikPrantare - you've refactored the beat detection code a while ago. Would you mind looking into this again and implement it in a more Milkdrop'esque way to keep the value ranges in the expected range?
The text was updated successfully, but these errors were encountered: