A bare-feet concept LoRA for MOSS-Image. Trigger word: feetTrained on 300 real studio photographs (watermark removal, resolutionbucketing, natural-language captions). It reliably produces the"both soles facing the camera, toes clearly defined" composition,and stays clean: without the trigger word it does NOT force feet intounrelated scenes.Recommended: er_sde / 25 steps / cfg 3.0 / empty negative / strength 0.9---## Tested base modelsWorks on all 9 MOSS bases I tested. Best results:| Base | Steps / CFG | Note ||---|---|---|| <code>moss_image_sft</code> | 25 / 3.0 | cleanest, official aesthetic || <code>Moss-dirtyMoss</code> | 25 / 3.0 | casual snapshot look, 8s || <code>Moss-artiusMoss</code> | 10 / 1.0 | fastest (4-6s), built-in turbo LoRA || <code>Moss-moodyMossMix</code> | 25 / 3.0 | best skin texture || <code>Moss-大师pornmasterMoss</code> | 25 / 3.0 | keeps western adult faces best |Avoid <code>Moss-动漫miaomiaoHaremMOSS</code> — body-shot composition tends to break.---## What it does / does not do✅ Verified- Close-up with soles facing camera- Both feet side by side, toes spread- Socks (half-off style appears correctly when prompted)- Full-body shots stay normal when <code>feet</code> is not written⚠️ Known limitation- Faces drift toward East-Asian / younger appearance. The training set camefrom one demographic, so that trait is learned along with the concept.This is not a plug-in that only adds feet. To suppress: strength 0.7 +face-lock prompt <code>caucasian woman, mature face, adult</code>).- Toenail polish appears only sometimes (56/300 in the dataset)❌ Don't expect- Style / lighting change (it's a concept LoRA, not a style LoRA)- Feet appearing without the trigger word---## Training details| Item | Value ||---|---|| Base | MOSS-Image Base (bf16) || Trainer | kohya sd-scripts <code>moss_image_train_network.py</code> || Dataset | 300 real photos, 4 buckets || Steps | 2400 (300 × repeat2 × 4 epochs) || Time | 1h20m @ 1.93 s/step on RTX 5090 || Network | dim 32 / alpha 32, <code>networks.lora_generic</code> || Timestep | truncated_logit_normal (logit_mean -1.09, logit_std 0.95) || Text encoder | frozen (official recommendation) |---## CompatibilityRequires a ComfyUI build with MOSS-Image support (0.42.0+ tested).<code>CLIPLoader</code> type must be <code>moss_image</code>. VAE is <code>flux2-vae</code>.The LoRA is in kohya format and loads directly — no conversion needed.

feet  moss

---## Tested base modelsWorks on all 9 MOSS bases I tested. Best results:| Base | Steps / CFG | Note ||---|---|---|| <code>moss_image_sft</code> | 25 / 3.0 | cleanest, official aesthetic || <code>Moss-dirtyMoss</code> | 25 / 3.0 | casual snapshot look, 8s || <code>Moss-artiusMoss</code> | 10 / 1.0 | fastest (4-6s), built-in turbo LoRA || <code>Moss-moodyMossMix</code> | 25 / 3.0 | best skin texture || <code>Moss-大师pornmasterMoss</code> | 25 / 3.0 | keeps western adult faces best |Avoid <code>Moss-动漫miaomiaoHaremMOSS</code> — body-shot composition tends to break.---## What it does / does not do✅ Verified- Close-up with soles facing camera- Both feet side by side, toes spread- Socks (half-off style appears correctly when prompted)- Full-body shots stay normal when <code>feet</code> is not written⚠️ Known limitation- Faces drift toward East-Asian / younger appearance. The training set came from one demographic, so that trait is learned along with the concept. This is not a plug-in that only adds feet. To suppress: strength 0.7 + face-lock prompt <code>caucasian woman, mature face, adult</code>).- Toenail polish appears only sometimes (56/300 in the dataset)❌ Don't expect- Style / lighting change (it's a concept LoRA, not a style LoRA)- Feet appearing without the trigger word---## Training details| Item | Value ||---|---|| Base | MOSS-Image Base (bf16) || Trainer | kohya sd-scripts <code>moss_image_train_network.py</code> || Dataset | 300 real photos, 4 buckets || Steps | 2400 (300 × repeat2 × 4 epochs) || Time | 1h20m @ 1.93 s/step on RTX 5090 || Network | dim 32 / alpha 32, <code>networks.lora_generic</code> || Timestep | truncated_logit_normal (logit_mean -1.09, logit_std 0.95) || Text encoder | frozen (official recommendation) |---## CompatibilityRequires a ComfyUI build with MOSS-Image support (0.42.0+ tested).<code>CLIPLoader</code> type must be <code>moss_image</code>. VAE is <code>flux2-vae</code>.The LoRA is in kohya format and loads directly — no conversion needed.

Files with SHA256: 91b420a75e73675e1bc5a8323c808996a228ef406968ddf35bdbe8ff7af0a5fd