Issue- Dynamic Link Breaks wile exporting comps in after effects 25.0 and above
Adobe After Effects version number - all the After effects 2025 builds has same issue
Operating system - windows 11
Steps to reproduce- open a sequence > select some clips in active seq > right click > "replace with after effects" > a dynamiclly linked comp apears in PPRO ,
PPRO is running in background ,Now when we render any comp inside after effects the dynamically linked files in PPro goes Offline
Expected result- export in after effects compleates without making PPro dynamically linked clips go offline
Actual result- dynamically linked files in PPro goes Offline
Possible issue-- the issue can be related to aerender.exe not mainaining the link with ppro wile exporting and making the clips in ppro go offline
Latest - "I found that this issue happens when a PPro project consist dynamically linked comps from more than 1 Ae project and looses link with the comps of the ae project where render was performed ....." also if same project is import in 2 diffrent ways like using dynamic link and another simplly importing the ae project file to premiere project , they are treated as 2 projects in one ppro project and thus the rendering ae comps causes aep files to go offline in ppro
I want to understanding GPU transform coordinate space , motion logic
I'm developing a CUDA GPU effect for Premiere Pro using the C++ SDK. The effect is essentially a custom transform/slide effect that translates the input image by sampling it with an offset (inverse mapping with bilinear interpolation).
I'm trying to match Premiere's built-in Motion/Transform behavior, but I'm running into something I can't explain.
The issue
The effect works correctly in many cases, but I consistently see this behavior:
Negative Y translations work perfectly.
If I set both Start Y and End Y to a negative value (for example -60), I can freely change the X position and everything still renders correctly.
However, as soon as the Y value becomes positive (even +1), the image immediately becomes distorted. It looks almost like the alpha and color data are no longer aligned, or the image is being sampled from the wrong rows.
I've already verified that:
Out-of-bounds samples return transparent pixels.
The bilinear sampler is clamped correctly.
Source width/height and pitch calculations have been checked.
The issue still persists.
My questions
What coordinate space does a Point parameter (mPoint) use in Premiere Pro GPU effects?
Is it pixel coordinates?
1/10 pixel units?
Normalized (0-1)?
Relative to the image center?
Relative to the top-left corner?
Does this differ between CPU and GPU rendering?
Is there an official SDK sample that demonstrates a GPU transform (translation/position) effect rather than a simple per-pixel effect?
How does Premiere's built-in Motion/Transform effect internally interpret Position values before they reach the renderer? Is there any documentation describing the coordinate conversion?
When implementing a GPU transform, is there a recommended way to convert Point parameters into pixel offsets?
Are GPU image buffers always top-left origin, or are there situations where the Y axis is inverted depending on the pixel format or rendering pipeline?
Is there anything special about Premiere's GPU frame layout (pitch, row alignment, coordinate system, or image origin) that needs to be considered when implementing an affine transform?
Has anyone implemented a GPU translation/transform effect with the Premiere SDK and can share the general approach (not proprietary code), particularly how Position values are converted into sampling coordinates?
I'm not looking for Adobe's internal implementation—I'm just trying to understand the expected coordinate space and the correct SDK workflow so my transform behaves the same way as Premiere's built-in Motion effect.
Any pointers to SDK documentation, sample projects, forum discussions, or best practices would be greatly appreciated.
AdChoices