Qwen Edit 2.1 fails at compositing characters into scenes — what model for film storyboards?
carmidian · reddit · 2026-11-01
The poster wants to build storyboards for a short film: placing multiple reference characters into specified locations with specified poses inside ComfyUI. Qwen Edit 2.1 produces poor results — it often just "photoshops" the character in, and even when body pose and lighting are right, the face stays frozen as the reference image staring at the camera. They're asking for better model and workflow recommendations.
More from Multimodal
- AI Video Gives George the Girl at the Dance, and Reddit Says the Timeline Is Cooked — jorgeecarpio · 2026-11-01
- "Off-axis perspective" AI image trend shows striking off-angle renders — XRarchitect · 2026-11-01
- AI-generated drama praised as indistinguishable from real footage — xiaosun86 · 2026-11-01
- Qwen Image Edit 2509 lands and the community calls it a huge upgrade — lesbwati · 2026-11-01
- AI-generated 3-episode horror series 'The Monkey's Paw' releases trailer — VastTVAI · 2026-11-01
- Blender previs + MiniMax H3 reference-to-video: a reproducible anime workflow — Time-Ad-7720 · 2026-11-01