Repository navigation
Releases: amariichi/LookingGlassGoViewer
Release list
v1.4.1
Changes
- Fixed a crop-edge depth artifact where transparent outside vertices could fall to zero depth and create a thin raised frame at the image edge.
- Updated Unity bundleVersion and README download links to v1.4.1.
Verification
- Confirmed in Unity by the user.
- Ran git diff --check on the changed files.
v1.4.0 release
GPU depth displacement and FOV metadata update.
- Move depth-based mesh vertex displacement to a GPU shader.
- Embed Depth Pro focal length and FOV metadata in RGBDE PNG files and use it for FOV correction in Unity.
- Add the FOV Strength slider for adjusting XY perspective spread separately from Z depth strength.
- Keep 15% image/depth overscan outside the visible frame for the 3D mesh to reduce visible cut edges when rotating.
- Optimize RGBDE splitting/cropping paths and update README / project version to v1.4.0.
v1.3.2 release
The script depth-pro_rgbde.py has been updated. OpenCV is no longer required, and it now works solely with Pillow.
v1.3.1 release
Update Unity Editor version and other packages version.
No substantial changes.
v1.3 release
Improve depth projection by reprojecting mesh vertices along camera FOV directions and linking slider ranges to HologramCamera settings.
v1.2 release
feat: enable arrow-key rotation (±30°); switch depth calc to linear
- Allow rotating the model up/down/left/right up to 30° with arrow keys.
- Change the model's depth calculation to a linear method.
v1.1.8 release
I simplified the texture scaling and cropping logic in ImageCropper.cs using openrouter/optimus-alpha.
Additionally, I made improvements to reduce visual artifacts ("edge noise") on the texture boundaries.
v1.1.7 release
Using openrouter/optimus-alpha, the code cleaned up. There are no changes to its functionality or behavior.
v1.1.6 release
I fixed the issue where the top and bottom of the vertically long image were being cut off.
I set the Hologram Camera's FOV to 40.
v1.1.5 release
I had GitHub Copilot improve the code in ImageCropper.cs, and since it worked when applied as is, I implemented it. The content remains unchanged. It might have run a little faster.