My attempt at creating a reference sheet model. Trained on 120 images which were mostly manually edited by me to remove stuff like signatures or watermarks. Model works great with the strength of 1. Trigger words are: reference sheet and model sheet
Recommendations
Steps: 32-37
CFG: 7
Resolution: 512x512, 512x768, 768x512
Prompt:
<lora:reference sheet-model sheet:1> , reference sheet, model sheet, simple background,
Or
<lora:reference sheet-model sheet:1> , reference sheet, model sheet, simple background, (multiple views, [full body|from behind|from side|from front|close-up|upper body|portrait]:1),
You can experiment with other types of background if you don't want it to be simple. Training images were mostly based on white/grey and black backgrounds.
Description
First Version
FAQ
Comments (7)
Just a heads up:
The first half of the example images have gone through img2img and inpainting. So re-using the same metadata will not result in the same image because the seed has been changed.
The second half of the example images have just went through hires.fix so re-using the same metadata should result in the same images.
You'll be able to tell the difference between these by looking at the difference in detail like the face/hands or clothes.
这真是我需要的，感谢！
Excuse me, if I only need one front and one back, how should I write Prompt
What about it?
pony version one day?
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.