ComfyUI beginner seeks advice: training LoRAs but struggling with pose control and backgrounds
hanosukee · reddit · 2026-08-25
A ComfyUI beginner with an RTX 3060 12GB shares their progress: they successfully trained a style LoRA using 20 images with Codex's help but hit a wall with pose control and background consistency.
Key Issues:
- Uncertainty about which base model (Anima, Pony, Illustrious) to commit to.
- Pose control nodes failing to make characters follow the intended poses accurately.
- Inconsistent background generation.
The user seeks recommendations on specific nodes, techniques, or workflows to master next in order to move beyond basic LoRA combining and prompt editing.
More from Multimodal
- MiniMax Releases H3-RAVEN Streaming LoRA for Real-time Video Generation — mvp-lab · 2026-08-25
- Localized Editing Might Be GenAI's Sweet Spot: Image Edit Test — Lonelydude014 · 2026-08-25
- Minimax H3 Text-to-Video Test: Dynamic Storm Background Prompt — azed_ai · 2026-08-25
- Wan 3.0 Turns a Dessert Cup Into a Flagship Ad Campaign — Prompt Breakdown — alifcoder · 2026-08-25
- Retro Model: Recalling ControlNet QR Code Monster — linoy_tsaban · 2026-08-25
- AI 'Stoneify Effect' generates stunning stone-textured images — bennash · 2026-08-25