Adding 3D Body Scans to LoRA Training
tekprodfx16 · reddit · 2026-07-14
The author forked AI Toolkit to add a second training stage based on SAM 3D Body, enabling LoRA/LoKr to learn not just faces, but accurate body shapes and proportions.
The workflow involves:
- Running standard LoRA/LoKr training to learn trigger words and reference photos
- Generating images and using SAM 3D to scan the bodies in the outputs
- Constraining the training with "real 3D body readings" instead of relying purely on pixels and captions
The author reports a significant improvement in full-body likeness, working seamlessly on Krea 2. Face training takes about 30 minutes, while adding the SAM 3D body stage extends it to roughly 60 minutes. The repo is FedorAiToolkit, and the SAM 3D Body model requires gated access on Hugging Face.
Related event: Enhancing LoRA Training with SAM 3D for Accurate Body Shapes(2 posts)→
More from Multimodal
- MOSS Image Edit vs Fibo-2.0-edit: 5-grid comparison of instruction-based image editing — Illustrious_Row_9971 · 2026-10-27
- Cinematic title sequence built with ComfyUI: LTX-2.5 + MiniMax H3 image-to-video — magratheya_64 · 2026-10-27
- Fish Audio ships open-weight Drama 3 TTS; creator shows turning old posts into narrated videos — alex_verem · 2026-10-27
- Redditor crafts photorealistic 'morning bus ride' short film entirely in ComfyUI — no3us · 2026-10-27
- New blur/anti-blur slider LoRA for MOSS lets you dial background sharpness with weights — NoMouse9610 · 2026-10-27
- PixGR: Pixel-space generative decoding for fast feed-forward 3D reconstruction — kwangmoo_yi · 2026-10-27