Skip to content

Update "Align Image to Reference" node to v3.0.0#3091

Merged
joeyballentine merged 3 commits intochaiNNer-org:mainfrom
pifroggi:main
Mar 12, 2025
Merged

Update "Align Image to Reference" node to v3.0.0#3091
joeyballentine merged 3 commits intochaiNNer-org:mainfrom
pifroggi:main

Conversation

@pifroggi
Copy link
Contributor

This brings the node up to date to match the Vapoursynth version.

Full changelog:

  • Overhaul to how the alignment works, which makes it easier to get good results. Previously it could be hard to finetune the options. Sometimes higher precision was better, sometimes multiple passes were better, sometimes more blur, sometimes a combination of all. The new version should now consistently improve with a higher precision level and no longer require much time to find the right combination.
  • Precision levels now produce different results. There are now 4 levels instead of 5. Alignment quality is better than before for a given precision level, but also slower.
  • Added new optional Mask input. It takes a black & white image that can be used to exclude areas in white from warping, like a watermark or text that is only on one image. Masked areas will instead be warped like the surroundings.
  • Added new Wide Search toggle and removed now obsolete Alignment Passes slider. This enables a larger search radius at the cost of some speed. When enabled completely different crops like 4:3 and 16:9, shearing, and rotations up to 45° can be aligned. It works by doing a fast homography alignment with the help of XFeat first to move the corners of the image roughly into the right position before proceeding with the main alignment step. This is both better and faster than the old approach.
  • Added new Low Quality Input toggle and removed now obsolete Blur Strength slider.
    This enables better handling for low-quality input images. When enabled general shapes are prioritized over high-frequency details like noise, grain, or compression artifacts by averaging the warping across a small area. This works better than the old approach and does not require finetuning.
  • Added fp16 support.

Copy link
Member

@joeyballentine joeyballentine left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Awesome, thanks. Heard this works really well

@joeyballentine joeyballentine merged commit 4207058 into chaiNNer-org:main Mar 12, 2025
14 checks passed
clausqr pushed a commit to clausqr/chaiNNer that referenced this pull request Jun 25, 2025
* Update "Align Image to Reference" node to v3.0.0

* Update "Align Image to Reference" node to v3.0.0

* fix pyright errors and improve description
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants