BR: Warp time

[fromwithin]

I've been using the Warp Time feature to correct some rhythmical errors in some acoustic guitar, but I've noticed that it's been weird at some points. Hard to describe what's wrong, but just not right.

I've been doing some investigating and it's a bit broken. The best way to test a lot of audio stuff is to use a sine wave, so I generated one to see what Warp Time would do to it.

First bug:
Each modification of the warp markers forces a re-render of the audio file. I added a few and shifted them around (we'll get to the results in a bit). On disabling the Warp Time feature, the audio clip did not revert to the original non-warped clip. Subsequent edits, like adding a speed change via time-stretching modified the warped clip, even though Warp Time was switched off. The only way I found to revert it was to close and re-open the project, or switch Warp Time back on, remove all the markers, then switch it off again.

Second bug:
The resultant audio at the time of each warp marker is pretty corrupted. In the attached images, I've set the warp markers and stretched each one slightly further than the last. The first image shows the zoomed-out view of the result. The second image shows a zoomed-in view of the 5th marker. It's a mess.

[AGreen]

Do you get similar results if you use warp markers in the clip effects rather than the loop properties window. I've always wondered why there are two places to add warp markers that seem to be independent of one another rather than just one. Surely just one would be better. Also when adding warp markers in clip effects there seems to be no way to choose the stretch algorithm or is that selected in the clip properties pane. Which algorithm is being used by default?

[fromwithin]

It's exactly the same if done as a clip effect instead of in Loop Properties. I didn't even realise that you could do it as a clip effect. If it worked properly that would be loads better for me.

You can't choose the stretch algorithm in the Loop Properties either. The algorithm option applies to a global time stretch, not to the Warp Time effect.

[chico.co.uk]

Loop properties warp time predates clip effects. I never used warp time when it was in loop properties, because i needed it to line up with other stuff in the edit, to correct my own timing mistakes, that kind of thing. And i found that a faff in loop properties. I use it quite a bit as a clip effect

[chico.co.uk]

Page 164 of the manual says you can choose the warp time stretch algorythm in the loop properties.

[AGreen]

Page 164 of the manual says you can choose the warp time stretch algorythm in the loop properties.

Thanks, looks like elastique pro is the default. Do you know if the clip effects algorithm is also set in loop properties.

[chico.co.uk]

I dunno tbh. Not had time to test really

[fromwithin]

Page 164 of the manual says you can choose the warp time stretch algorythm in the loop properties.

I've tried all of the stretch modes in Warp Time and they don't do anything to the audio when you change to a different one. Maybe that's another bug related to it not re-rendering the audio when you disable Warp Time.

[AGreen]

Have you tried using Groove Doctor to correct timing issues using the spliting mode (no time stretch). I always try that before resorting to time stretch because time stretch does tend to introduce artifacts if you listen carefully. I recently corrected some timing on a recorded bass guitar using that method and it sounded excellent, completely transparent.

[fromwithin]

Just tried Groove Doctor and it's just as bad.

I tried it with the acoustic guitar track and the results were dreadful. All clicky and nasty. To try to hear what was going on, I analyzed the acoustic guitar to get the split points and then replaced the clip with a sine wave clip. Then I had the Groove Doctor operate on the Sine wave instead of the acoustic guitar.

Here's the awful result...

https://app.box.com/s/lhmpd16jjbj8wk58z46m4kj4v7xaorpa

[AGreen]

It will mangle a sine wave because it's splitting and moving parts with no time stretch and with padding offsets etc. It's not meant to produce a perfect stretched copy of the part, it's meant to produce one that sounds ok. If the part is too busy or too uniform (a sine wave is extremely uniform) you may get double transients or repeated audio which will sound wrong. Messing around with the crossfade and offset values can produce good results with real audio and a little patience.

[fromwithin]

The Groove Doctor operation I was doing was "Warp quantise/groove", which appears to operate exactly like Warp Time.

Of course it will never be perfect with a sine wave, but it should never be producing the sharp click points in the second image I posted (in the first post). Also, presuming that there is some kind of crossfade between each transition point I'm not sure that the crossfade method being used is the right one. I don't know exactly what it's doing right now, but It seems like it's doing some kind of equal-power crossfade. I tried it on a DC waveform and you can see the weird shape that appears at every transition point. As the waveforms at the transition points are going to be near-correlated, I would expect it to be an equal-voltage crossfade.

I might also be completely misinterpreting what's going on. Only the devs will know.

[AGreen]

I'm talking about the Quantise/groove mode which doesn't warp audio at all, it splits and quantises the clip positions. It won't perform well on a sine wave because you can't split and move bits of a sine wave around without producing gaps or overlapping audio, but it can work very well on certain types of audio. As I said I got very good results correcting a bass part with it and even some strummed guitar parts have worked ok. It's very dependent on the audio.