Beta version 23.04 now available to test

Windows 11 - 22H2

Shotcut V23.04.14

I have the same problem as MusicalBox with .mp4 files

Crash with Speed: Forward & Reverse and also with Time Remap filters, as soon as click OK in the conversion window, with Good, Better and Best formats.

Tested with .mov and .avi : No need for edit-friendly conversion.

Installed on Manjaro KDE wishing to test colour correction with some 10-bit footage. Found the GPU Sharpen filter goes from unsharpened to vastly oversharpened with the minutest movement of the sliders, making it impossible to find a sweet spot.

In case this can be helpful:

  • I noticed that right after clicking on Save, Shotcut’s memory usage drops considerably before the crash.

  • Also here’s the Application Log, right after the crash: shotcut-log.txt (36.0 KB)

I saw this on GitHub

So I downloaded the latest nightly build (v23.04.16), and so far all the files I tried were converted successfully. From the Speed: Forward & Reverse and the Time Remap filters.

:+1:

2 Likes

That’s great. I did not reproduce it, so I followed the code path and made an educated guess.

1 Like

Filter: Simple Shape Mask
I have a problem with keyframes.
The “rotation” keys are disabled by themselves, and the position keys do not move the mouse around the timeline.

Windows 11 - 22H2
Shotcut V23.04.16

Export with h264_nvenc hardware encoder

The export .mp4 with the h264_nvenc encoder gives a bad result if the export does not have the same number of pfs as the project. (example: project 1920x1080 30fps and export 1920x1080 25fps… or 50 or 24…)
hevc_nvenc, h264_amf work fine.
It work with v22.12 and older versions.
See file properties:

fixed

1 Like

About the Motion Tracker filter.

Not bug to report at the moment, but I have a suggestion, if you don’t mind.

At the moment it generates a keyframe every 5 frames. In most cases this is more than enough. But in other situations, I feel that more keyframes would be necessary. When the object is moving quickly and makes sudden trajectory changes for example. If one of these trajectory changes happens between two keyframes, for a fraction of a second the tracker loses the object.
If you look closely at the exported clip in my post above, at some point the boxer makes a small jump and the text doesn’t quite follow her head during the jump.

That’s why I wonder if you would consider the idea of adding a setting that would allow the user to choose the number of keyframes generated by the filter. A drop down menu perhaps, or a spin box set by default to generate a keyframe every 5 frames, but allowing to change this value to others, ranging from 1 to 10 perhaps?

It is likely that generating a keyframe every 1 or 2 frames would involve a much longer analysis time and require more system resources. I think that even though these small values would rarely be used, they would be welcomed when better precision is needed.

On the other hand, when the object has a constant velocity and trajectory, the larger values would be sufficiently accurate and the analysis time would be shortened.

1 Like

If you add a new track and try to add a LUT (3D) filter to it, the program crashes.

With GPU Effects turned on adding any filter to an empty timeline is a problem that should be avoided for now. This is fixed for the next version.

1 Like

I just noticed even with the new QT version Shotcut still doesn’t follow changing system audio outputs, is this still bugged in QT or just wasn’t a priority so far?

Shotcut uses SDL, not Qt, for audio output. Each time we upgrade the SDL2.dll to fix this, we get reports from a minority that audio plays with awful distortion that goes away when provided the old version of that DLL. So, it gets reverted, and there it still stays.

Oh, didn’t know that.
Would maybe a workaround with some new menu entry (like say “Reload output”) in the Settings → Audio channels make sense? I tend to switch between speakers and headphones multiple times a day and it always takes me by surprise.

A workaround today is to make a small 1ms change in Settings > Player > Synchronization… But it does not make sense to place an explicit workaround in Settings. I will use Qt to detect the change to the audio output device and reload.

  1. To add to my earlier report in this thread about Blur: Gaussian:

It turns out that it’s also an issue with Blur: Box. However, Blur: Exponential and Blur: Low Pass are fine. Here are pictures to demonstrate this:

To compare here are the other two Blur filters:

And here you can see that the Gaussian and Box filters can blur far more without the Mask filters:

All of this happens with or without GPU on.

  1. There’s a strange flickering going on when moving a video or image with GPU Size & Position that doesn’t happen with the CPU one. Video demo.

  2. And to elaborate on my suggestion from before about this:

Giving this a second thought, instead of a few buttons for text and color clips, I would suggest that it be just one Generator button that pulls up only the list of Generator options from “Open Other” since all the items on that list can be added to the timeline. Also, since it would be on the timeline toolbar then “Add To Timeline” can be removed and “OK” or “Open” would just automatically place it on the timeline.

You can stop reporting about that. These blurs are not going to be changed.

I would suggest that it be just one Generator button

This too. Not going into this release.

You could’ve told me that the first time I reported on it so that I didn’t have to continue wasting time testing something that you had no intention of doing anything about.

Closing this thread now as there will be a new beta release in a day or two with its own thread.