This is particularly troubling when creating tutorials using the sequencer. If you do any zoom ins on menu items or text, it's quite hard to recognise what you're looking at.

In T78509#973685, @Evan Wilson (EAW) wrote:

Please, can you write your settings?
If we fix cryptomatte in blender out, we will make serial in it, its very important for us. Thank you!

I was traveling without my computer, so I couldn't tell you until now.
RGBA, color depth full, zip, save buffer no, out in render animation are my settings. When I was testing this days ago, I was seeing an image like yours, with only 2 colors. Today I tried a few things with your file, and I think the issue is that you are using the compositor in Blender with nothing attached from the Render Layers node to the Compositor Image output. Also the Alpha channel isn't connected, and Use Alpha boxes aren't checked in your file.

Attaching the Image, Alpha, and checking the box makes it work for me in Nuke.

As this is the Blender bug tracker, I will close this report since the issue with compression is a cryptomatte issue and not a Blender specific one. I will test the save buffers issue a little longer, then will open a new report about it and tag you in it.

Please, can you write your settings?
If we fix cryptomatte in blender out, we will make serial in it, its very important for us. Thank you!

In T54273#939936, @Guillem de Diego (Unsigno) wrote:

Then it should be changed to "feature request", but working with images on video editing is a common requirement and the current approach makes it hard and counterintuitive.

I tryed render like your say, cryptomatte in no changed.
Find only metadata. Crtmt is broken.

Cryptomatte and metadata loss if file have compression.

We try to work with instruction od updated psyop. It isnt Nuke's trouble. Blender is dont write data in every time.
Shortly - yes, we sure (command of film developers.)

@Sergey Sharybin (sergey) I made the new report: https://developer.blender.org/T78579

Please make a new report, and attach all the requested information (.blend, blender version and so on). Makes it easier to re-iterate over the fix.

I just got this bug in 2.90. Should I create a new task with the detailed steps to reproduce or put everything in here?

Yes in 2.9

Sorry, just to confirm. Is it solved in Blender 2.9? For I saw the issue in 2.83.
Thanks for letting us know

Should be fixed in latest 2.90 builds https://builder.blender.org/download/

Anyone knows workaround for this? 2.83.1 still do not work. How to get that Z channel thing to work?

In T77671#964370, @Sergey Sharybin (sergey) wrote:

Could also be related on "Save property value for the next operator run". In this case "just" assign all operator properties in the keymap and see if that helps.

It looks like I accidentally fixed this with rB8bcfa0d1fb9f. Please let me know if the issue still persists.
Sorry for this taking so much time. It was a bit convoluted problem.

In T77564#953852, @Jeroen Bakker (jbakker) wrote:

My proposal is to make the scene sequencer preview a GPU viewport this is the most future proof way and use the global window stereo setting.
Current implementation seems to be a per strip setting. But that will change current workflow.

@Richard Antalik (ISS): What do you think?

With image data, the results are still different. We have prepared the data.
This is a lightweight sequence data converted to JPEG, but it has a FPS drop.

Will classify as bug for now.

Changing the handle type seems to play an important part, animation of just the point positions seems to be fine.

in 2.90 I get
65FPS without background image
25FPS with background image

In T77023#964394, @Sergey Sharybin (sergey) wrote:

@Richard Antalik (ISS), from reading the discussion seems this is a an out-of-memory issue?

Not sure if rB5cfbc722d095: Fix T78047: Fix failing denoiser tests on windows possibly changed this for the better?

Thanks!!! Will this be ported to 2.83 as well? It seems to be a pretty straightforward and non code breaking patch.

I have this on the list for tomorrow to look at again

In T76015#964367, @Sergey Sharybin (sergey) wrote:

@Philipp Oeser (lichtwerk) I'm a bit confused here. To my knowledge the module project is assigned once the issue is reproduced and confirmed. AM I missing something, do we have confirmed "repro" here or is this still needs triaging?

@Richard Antalik (ISS), I do have some FPS drop, but seems to be within expected threshold. How much FPS loss happens in your case?

@Philipp Oeser (lichtwerk) mask surely shouldn't be mangled after modifying it.
Could be missing DEG_id_update_tag call. How do i reproduce the issue with the attached file?

@Richard Antalik (ISS), from reading the discussion seems this is a an out-of-memory issue?

No activity for more than a week. As per the tracker policy we assume the issue is gone and can be closed.

Could also be related on "Save property value for the next operator run". In this case "just" assign all operator properties in the keymap and see if that helps.

@Philipp Oeser (lichtwerk) I'm a bit confused here. To my knowledge the module project is assigned once the issue is reproduced and confirmed. AM I missing something, do we have confirmed "repro" here or is this still needs triaging?

i'm getting the same thing. https://developer.blender.org/T78200

There are two aspects here:

In T78129#964061, @Sergey Sharybin (sergey) wrote:

Quick guess: pasting strip should tag depsgraph relations for update.

reminds me of T69430

I can open file though in movie clip editor, so issue will be on our side, which I suspect would be the case. I can look into this. Should be pretty straight forward to track how the string is processed.

Quick guess: pasting strip should tag depsgraph relations for update.

Hmm filename is messed up by phabricator. But if I change file name to C​a​l​l​ ​o​f​ ​D​u​t​y​®​_​ ​M​o​d​e​r​n​ ​W​a​r​f​a​r​e​®​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​ 2020-06-15 02-33-57.MP4 I can reproduce.

@Richard Antalik (ISS), do you mind trying to download file and reproduce it again? Just to be sure phabricator doesn't "sanitize" the name in some way. Also, can you share your exact steps, maybe I'm missing something.
For me the file opens fine in both Movie Clip and Sequencer editors.

@Sergey Sharybin (sergey) yes, I was able to reproduce with following file: F8640318

@Richard Antalik (ISS), are you able to reproduce the issue? I've tried renaming a lot of files here and everything still works correct.

I would simply close this task, I do not consider this to be an issue to be fixed at all. Blender should not track the file system for auto updating.

@Philipp Oeser (lichtwerk), design task is for design proposals, with concrete outline what and how is to to be implemented. Design limitations do not really fit under design tasks, this is a known issue.

Here is an ASAN output: P1482.

I have been able to reproduce from within debugger, but only when I run with -d argument.
So far I am not sure if this is bug in tool system in general or in VSE implementation.

Thanks for crash file. I got confused and thought that crash was on linux.

In T78112#963487, @Christopher_Anderssarian wrote:

Could you run the debug argument with 2.90.0, there should be a stack trace attached to that.

Could you run the debug argument with 2.90.0, there should be a stack trace attached to that.

In T78112#962770, @Richard Antalik (ISS) wrote:

In T78112#961815, @Paulo Suzuki (pedropaulosuzuki) wrote:

In T78112#961806, @Richard Antalik (ISS) wrote:

Can you update your GPU driver and check if this resolves the problem?

Just updated to latest version and the issue still remains.

Driver 391.35 for GT 630M on Windows 10 Home. Not an issue on Ubuntu 20.04 LTS, will check driver version there as well.

I suspect some openGL issue here not necessarily linked to VSE exclusively.

Can you run Blender from command line with -d argument? I gues it would be blender -d and paste here console output? I am not sure if it will contain stack trace in Linux, if not there will be crash.txt file mentioned in the end please upload that as well.

I assume this will be issue in movie loading code somewhere since sound is working correctly.