[Feature request] A (better) way to assign sample slices to notes

[linux-leo]

I recently wanted to make breakcore/ambient-jungle type music in radium. To get precise cuts, I used propellerhead recycle to slice the amen break into 45 individual hits. And wanted to play those mapped to individual notes. However, every approach and alternative I could think of so far was flawed in some way or another.

1. Convert to Soundfont and play in the Sample Player: Radium doesn't seem to recognise when I set a fixed key for all drum hits so that they are played at the same pitch, instead the drum hits are pitched according to the notes they are mapped to. Furthermore, every attempt at creating a soundfont with polyphone or swami (regardless of which samples I used or if it was a melodic or percussion instrument) failed to play properly in the radium sample player, either producing bad resampling/aliasing artifacts or sounding way to muffled, depending on which resampling algorithm I chose. Maybe I just suck at creating soundfonts, since many of my downloaded soundfonts work fine in the Sample Player, however the soundfonts I create myself play fine in Fluidsynth. Also the stereo channels are not displayed side by side like when the samples are played normally.

2. Convert to Soundfont and play in Fluidsynth: works, but again no visualization in the editor view, which is important for lining up the hits imo, and I can't do pitch bends, portamento, reverse sample playback like with the Sample Player.

3. Convert to .xi: The .xi Format has two limitations which suck, one it only supports mono samples, two it supports by convention only up to 16 samples per instrument. I think I can work around the latter by manually adding more samples in a hex editor and around the former by mapping left and right channels a certain distance of semitones apart and triggering them at the same time with the note duplicator settings at the left. However both channels still won't be displayed in the editor.

4. Slice in Radium with automation instead. It unfortunately is not as accurate as using external software for automatic slicing. And I need to both trigger automation and the sample whenever I want to play a slice.

5. Use the sequencer instead: One of the best options, but it is very time consuming copying, pasting and moving the same samples over and over compared to simply entering a note. I also end up with a huge amount of samples, which can crash radium occasionally.

Sorry if this comes across as a bit picky. But every method I've tried has downsides. So either adding another method to map slices to notes, or improving one of the existing ones would be much appreciated.

[linux-leo]

One option is adding ".iti" IT instrument support to the sample player. It supports stereo samples (mentioned at the end of https://wiki.openmpt.org/Development:_Formats/IT) and can store much more samples than .xi, it also wouldn't require adding any more UI code. Instead delegating sample mapping to one of the many free .IT compatible trackers.

[kmatheussen]

Can you send me a soundfont that doesn't play correctly? Also, is the muffled sound be caused by resampling algorithm, or is something else at play? Is it muffled even when using SINC resampling? I don't remember what it's called, but I'm pretty sure fluidsynth uses a resampling algorithm not supported in Radium, but I've been thinking about including it for a long time, it shouldn't be very hard, I think it's even available in JUCE.

Regarding "Also the stereo channels are not displayed side by side like when the samples are played normally.", do you mean that wrong samples are displayed in the editor? E.g. that sample #0 is always displayed, not the one(s) that is actually played for the note? (I'm just guessing here, I don't know what you mean.)

".iti" might not be a lot of work to implement if just ignoring things like sustain-loop, etc., but I rather want to make soundfonts play correctly first.

[linux-leo]

I took another look at the issue and replicated the problems I had and tried to think it better through. Sorry for leaving this very confusing rant type issue, I wasn't thinking straight.

First I created a soundfont with swami, creating a percussion preset, then assigning an instrument to it, and then adding left and right samples of the first slice to C3 and the left and right samples of the second slice to C#3 and setting the root note to 60 and fixed key to 60, so that the sample is played back unpitched, C3 and C#3 map to C4 and C#4 in Radium however, not sure why. And it appears that the fixed key setting is not respected, therefore pitching the sample down from C4 to C3, despite being mapped to C4 in Radium (???). The weird resampling artifacts are however not just a side effect of repitching from C4 to C3, as pitching a regular sample up and down sounds much better, regardless of the resampling algorithm used. Maybe there is some numerical inaccuracy regarding pitch calculation.

the stereo channels are not displayed side by side like when the samples are played normally.

Ignore that last part, after creating my soundfont again and it looks correct. I must have misremembered or made a mistake creating the soundfont the first time.

Attached below is a radium project embedding the soundfont, playing it first in fluidsynth, then in the sample player.

test.zip

[linux-leo]

Setting the relative note instead of specifying a fixed key works, but for some reason, radium plays it one octave lower than the specified note. And if I set the relative note one octave below the assigned key, then radium plays it at the correct speed.

[linux-leo]

For what it's worth, here is a second version with again both fluidsynth and the sample player playing the same soundfont, except with the root note set to the assigned key instead of the fixed key set to the root note, and with the root note adjusted to be one octave below the assigned key.

testtwo.zip

Besides the resampling inconsistencies I would say the issue boils down to radium playing a soundfont one octave lower than fluidsynth and not respecting the fixed key setting. I found workarounds to both though (using root note instead of fixed key and setting it an octave lower) by replicating this issue again. And resampling arrifacts dissappear when pitching up instead of down, which is what you do for jungle/breakcore etc... anyway, so I'll keep working on my projects for now.

[kmatheussen]

Thanks for investigating. This is really bad. Will fix.

[linux-leo]

I can confirm the patch you provided works and improves soundfont playback, making it more consistent with fluidsynth.