Analysis waveform drifts when exporting metadata after analyzing an MP3 file with ID3 v1.1 tags #11159
Comments
|
Caused by a missing Getting rid of |
|
This issue affects all MP3 files that contain any MP3 metadata frames (ID3v1/ID3v2/APE) before the MP3 audio frames. |
|
I cannot confirm your assumptions. I am currently working on a solution with SoundSourceMp3. |
The fix is easy and obvious but it has serious implications. Nevertheless, the current situation is also bad. |
Great find! I was confident that something like this was the case considering the circumstances, but I hadn't looked at any code yet. If I don't forget to I'll try to force |
|
The fix seems to me more complicated, tests are failing. Nevertheless, the situation and its implications are as bad as they could be. The length and offset of the decoded audio when using |
|
Only the length changes on presence/absence of tags. The issue of skipping the decoding of the first frame has been introduced in SoundSourceMP3 in Mixxx 2.1.0 2015 #411 a fix workaround for non accurate seeks compared to start the track from the very beginning. |
|
Probably skipping the first frame is correct. I have recorded a wav file and a mp3 file with Mixxx the wav file starts immediately while the mp3 file has 20 ms noise before the signal starts. The skipped frame is completely silence. The issue might be that this frame is skipped unconditionally in Mixxx. Audacity decodes the frame as audio. |
|
#11168 is done now please have a look. |
|
I gave #11168 a brief try and it seems to fix the issue. Haven't done any extensive testing nor have I looked at the patch though. |
Bug Description
See the Zulip chat for more context: https://mixxx.zulipchat.com/#narrow/stream/109171-development/topic/Analysis.20waveform.20sometimes.20drifting.20slightly This one is super specific and it took a while to figure out a pattern. I'm still not entirely sure about why what's happening is causing problems, but it seemed like a good idea to at least submit a bug report for this now that we have a lead.
The problem is that when analyzing certain MP3 files and then saving the metadata to file tags (either because the relevant option in Settings -> Library is enabled or through Right click -> Metadata -> Export to file tags, this may require restarting Mixxx if the export is deferred), the analyzed waveform is incorrect. The waveform usually drifts to the left, although I've also seen the same thing happen with waveforms drifting to the right. This makes it look like the tempo is slightly off as the beat grid doesn't align with the waveform, even though the beat grid is actually correct. Using the scratching controls with a controller or the waveform display to focus on a transient will confirm that the beat grid is indeed correct and that the waveform display is off. If you then go to Right click -> Reset -> Waveform and then analyze the file again the waveform will be correct.
I had a hard time figuring out a pattern, but I just noticed that this only seems to happen with MP3 files containing ID3v1.1 tags in addition to ID3v2.4.0 tags. MP3 files containing just those ID3v2.4.0 tags don't have the issue.
This .zip archive contains two MP3 files: mp3-test-files.zip
think-break-with-ID3-tags.mp3is unaffected by the issue.think-break-with-additional-ID3-v1.1-tags.mp3is the exact same file, except for some additional ID3v1.1 tags added using kid3.Version
main branch, 86a280b
OS
Arch Linux
The text was updated successfully, but these errors were encountered: