About this Lora:
This Lora will create side by side images creating the illusion of stereoscopic vision by creating left and right images of either reference images.
Where it shines:
• Aspect ratios: 1:1 / 3:4 / 4:3 / 9:16
• close-up shots and Portraits
• realistic images / Photos (pretty decent at some 2d Styles)
What it struggles with:
• Aspect ratios: Anything exceeding 4:3 (i.e 16:9 and widescreen)
• Completely black areas that are supposed to be black objects
• Strongly postprocessed unrealistic photos
Testing:
• Pretty stable at a range from 2.00 - 2.50 (strength)
Caution:
This Lora was trained on NSFW Stereoscopic-3D Images.
Be aware that cranking up the Lora's strength can result in naked
body horror surprises (not the pleasant kind).
Promt advise:
SBS_Stereo_img SBS_Stereo_img.
stereogram.
SBS_Stereo_img SBS_Stereo_img.
stereogram.
Two images side by side with the exact same width.
On the left: image1 (keep the image1 exactly the same).
On the right: slightly shifted second image with slightly rotated perspective.
Example Workflow:
https://civarchive.com/models/2720656/stereoscopic-sbs-3d-image-workflow-flux2-klein-9b
Further Versions:
I might continue training, as i have some ideas how to improve this thing.
But still, the way this Lora works feels like a workaround, as it creates a
sbs-Image instead of an same-sized image for the other eye.
The way I trained it right now requires a workflow handling the additional
image space needed.
How did you train your sbs Loras? Please contact me.