Hips Shape Slider – LoRA for MOSS Image
A small concept slider for MOSS Image that makes the hips of a clothed adult figure wider as you raise the strength. One file, no trigger word. It is strongest on fitted jeans seen from behind and weak or invisible on loose or long clothing, so read the limits below before you download.
🎯 What it does (honest summary)
Same prompt and seed, MOSS Image SFT, 50 steps, with and without the LoRA. In our fitted-jeans, back-view example the hips get steadily wider from strength 0.5 to 1.5. In the other scenes we tried (straight trousers, long coat, long skirt, knee-length dresses, front-facing portraits) the change is small or not visible. In our own check of 8 prompts x 2 seeds at strengths 0 / 0.5 / 1.0 / 1.5, the hip width grew in the expected order in only about 5 to 6 of 16 pairs. We tested positive strengths only (0 to 1.5), not negative.
🖼️ Base vs LoRA
Same prompt and seed. Left: MOSS Image without the LoRA. Right: with this LoRA at the strength shown. The two jeans pictures at the top show the clearest effect we found; the last two are subtle and also show small changes in props or pose.



🚀 Quick start
Base model: MOSS Image (trained and tested on MOSS Image SFT, int8)
Trigger word: none
Strength: 1.0 to 1.5 (0.5 is a light touch). We did not test above 1.5 for release
Settings used for the examples: SFT, 50 steps, CFG 3, euler / sgm_uniform, shift 5, 896x1344
Prompt tip: say what the figure wears, for example fitted jeans or a fitted dress, and keep the framing from behind or full length
⚠️ Limits you should know
Little or no visible effect on long coats, straight-leg trousers and long loose skirts in our tests
At strength 1.0 and above, the background, pose, shoes or small props can change from the base picture (we saw a turned head, different footwear and a belt appear)
Tested on 8 prompts with 2 seeds each, all adults in everyday clothes. We did not test other styles, group pictures or other base models
There is already another hips slider for MOSS Image by another creator. We did not compare the two
✨ Features
One small safetensors file, about 71 MB, no trigger word
Slider behaviour: strength 0 leaves the picture unchanged
🧪 How it was made
Trained locally with ai-toolkit (concept slider) on adult pictures that I generated myself with MOSS Image. No photos of real people were used. Checkpoint at step 300 of the second training recipe. All example images are AI-generated.
⭐ Found this useful?
Like if it saved you time
Comment your results — I read every one
Follow for new ComfyUI workflows, all tested on 16 GB VRAM
⚖️ Attribution and licence
Base model: MOSS-Image by OpenMOSS-Team, weights released under Apache-2.0 (model card). This LoRA is a modification trained on top of it
Trained with ai-toolkit by Ostris

