LoRA Training with SAM 3D for Accurate Full-Body Proportions
tekprodfx16 · reddit · 2026-07-14
A developer forked the AI Toolkit project and integrated Meta's SAM 3D Body model, solving the issue of body type drift (height, proportions) during standard LoRA training for full-body generations.
Training Mechanism:
- Phase 1 (200 steps): Standard LoRA/LoKr training based solely on reference photos.
- Phase 2 (15-60 steps): After image generation, SAM 3D scans the body types of both generated and reference images, using a loss function to force the LoRA to learn the real person's 3D physical characteristics (rather than just guessing from text descriptions).
Performance & Compatibility:
- On an RTX 5090, training just the face takes 30 minutes; adding 3D body scanning takes 60 minutes.
- Currently supports the Krea 2 model; Ideogram 4 support is in development.
The author has open-sourced the project, significantly improving the realism and consistency of full-body character generations in AI art.
Related event: Enhancing LoRA Training with SAM 3D for Accurate Body Shapes(2 posts)→
More from Multimodal
- Wan2GP lands on Pinokio: 1-click AI video on 6GB GPUs — cocktailpeanut · 2026-10-27
- WanGP v18.81 adds Tencent's Hy-Nomas world model, playable on 8GB VRAM — cocktailpeanut · 2026-10-27
- Butter.video wins over skeptical creator with hundreds of tasteful timeline 'blocks' for video effects — beechinour · 2026-10-27
- AI Video Pits Brock Lesnar Against Duncan the Tall in a Toy Showdown — KloneStudios · 2026-10-27
- $50, 2-hour Kurzgesagt-style AI video reignites debate over intent in AI creation — motionbynick · 2026-10-27
- Fake BTS of a dam disaster made with Dreamina Seedance for 29 cents — Standard_Spell5873 · 2026-10-27