Audaspace: faster audio resampling #116059

Aras Pranckevicius · 2023-12-11T19:40:00+01:00

2023-12-11 19:40:00 +01:00

Previously in Audaspace there was choice between linear resampler (okay for preview, but not great for final mix), or "extremely high quality preset" for rendering the final mix; with nothing in between. I have just landed "medium" and "low" resampler quality levels in upstream Audaspace (see https://github.com/audaspace/audaspace/pull/18 with details and quality spectograms, also comparison with Audacity resampler).

This PR updates Audaspace to latest upstream, and switches to use the newly added "medium" quality resampler. There's no audible difference (nor visible one in spectrograms), as far as I can tell.

Timings, rendering out frames 1000-3000 of Sprite Fright Edit blender studio data set:

Windows (Ryzen 5950X, VS2022): 92 -> 73 sec
Mac (M1 Max, clang 15): 70 -> 62 sec

i.e. using a faster audio resampler makes the whole render process be 10-20% faster (however, this from VSE where it combines already pre-rendered image strips).

Previously in Audaspace there was choice between linear resampler (okay for preview, but not great for final mix), or "extremely high quality preset" for rendering the final mix; with nothing in between. I have just landed "medium" and "low" resampler quality levels in upstream Audaspace (see https://github.com/audaspace/audaspace/pull/18 with details and quality spectograms, also comparison with Audacity resampler). This PR updates Audaspace to latest upstream, and switches to use the newly added "medium" quality resampler. There's no audible difference (nor visible one in spectrograms), as far as I can tell. Timings, rendering out frames 1000-3000 of Sprite Fright Edit blender studio data set: - Windows (Ryzen 5950X, VS2022): 92 -> 73 sec - Mac (M1 Max, clang 15): 70 -> 62 sec i.e. using a faster audio resampler makes the _whole render process_ be 10-20% faster (however, this from VSE where it combines already pre-rendered image strips).

Aras Pranckevicius added this to the Video Sequencer project 2023-12-12 06:00:19 +01:00

Aras Pranckevicius force-pushed audaspace-resampler-opt from 4dbd9e776d to 50e04fe387

2024-01-02 20:27:58 +01:00

Compare

Aras Pranckevicius changed title from ~~WIP: Audaspace: faster audio resampling~~ to Audaspace: faster audio resampling

2024-01-02 20:57:28 +01:00

Aras Pranckevicius commented

2024-01-02 20:58:06 +01:00

@blender-bot build

Aras Pranckevicius requested review from Sergey Sharybin 2024-01-02 21:30:29 +01:00

Sergey Sharybin commented

2024-01-03 16:47:32 +01:00

Nice work!

Is it possible to set resampler quality to medium via some API instead of modifying the bundled sources? That'd make it easier to update the Audaspace in the future, and will make the update process more robust (so that we wouldn't loose the speedup by accident).

Nice work! Is it possible to set resampler quality to medium via some API instead of modifying the bundled sources? That'd make it easier to update the Audaspace in the future, and will make the update process more robust (so that we wouldn't loose the speedup by accident).

Aras Pranckevicius commented

2024-01-03 16:58:51 +01:00

Is it possible to set resampler quality to medium via some API instead of modifying the bundled sources? That'd make it easier to update the Audaspace in the future, and will make the update process more robust

I'd need to add said API to Audaspace upstream first, for that to happen. Should I try doing that?

> Is it possible to set resampler quality to medium via some API instead of modifying the bundled sources? That'd make it easier to update the Audaspace in the future, and will make the update process more robust I'd need to add said API to Audaspace upstream first, for that to happen. Should I try doing that?

Sergey Sharybin approved these changes 2024-01-03 17:27:47 +01:00

Sergey Sharybin left a comment

Ideally yes, that'd be very lovely. Will also make it so people who use stand-alone library (it seems it is packaged for Arch, i.e.) can benefit from the performance without re-compiling library.

That being said, I don't really mind if we commit the current patch as-is, as there is nothing better we can do with the current state of Audaspace API.

Not sure what level of review you expect btw. I can't test it beyond of how you've tested it, and the code has been reviewed by the Jeorg already in the Audaspace repository. So I'd trust your testing :)

Ideally yes, that'd be very lovely. Will also make it so people who use stand-alone library (it seems it is packaged for Arch, i.e.) can benefit from the performance without re-compiling library. That being said, I don't really mind if we commit the current patch as-is, as there is nothing better we can do with the current state of Audaspace API. Not sure what level of review you expect btw. I can't test it beyond of how you've tested it, and the code has been reviewed by the Jeorg already in the Audaspace repository. So I'd trust your testing :)

Aras Pranckevicius merged commit 986f493812 into main

2024-01-03 21:23:34 +01:00

Aras Pranckevicius referenced this issue from a commit

2024-01-03 21:23:36 +01:00

Audaspace: faster audio resampling

Aras Pranckevicius deleted branch audaspace-resampler-opt

2024-01-03 21:23:37 +01:00

Aras Pranckevicius commented

2024-01-04 14:29:06 +01:00

Alright, I've submitted PR to add API to upstream Audaspace https://github.com/audaspace/audaspace/pull/19 and once/if that is accepted then will apply to Blender codebase as well.

Arye Ramaty referenced this issue from a commit

2024-01-21 15:37:38 +01:00

Audaspace: faster audio resampling

Sign in to join this conversation.

No reviewers

No Label

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

Audaspace: faster audio resampling #116059