Qwen Image 2.1 nails 4/44-panel layouts in single-prompt character sheets, fails framing review
HarnessLightnin · reddit · 2026-09-21
A game art lab tested whether Qwen Image 2.1 (7B DiT) could replace their multi-stage character turnaround pipeline (H3 video turnaround + pose extraction + stitching) with a single-pass 4-panel composite prompt. Setup: single RTX 3090, 40 steps Euler/Simple, CFG 5.0, 46.7s per sheet, 14 GiB peak VRAM, 44-character batch; layout and 1968 studio-portrait style held constant while wardrobe/chassis varied, with print styling pushed to deterministic post-processing.
What worked (44/44): every sheet placed views in the correct quadrants with zero panel swaps; costume silhouettes, palettes, and materials stayed coherent across all four views without reference images; 45s per composite beats the multi-stage workflow.
What failed human exit review: dwarfism-like compressed proportions on some figures; inter-panel camera distance drift (side profile framed closer, vertically squashing anatomy); and full-body bound violations — same sheet cropping at shins in rear view and mid-thigh in side view.
Conclusion: prompt-only Qwen Image 2.1 reliably composes and keeps identity coherent but can't control per-panel camera distance or framing bounds — not yet a pipeline replacement.
More from Multimodal
- Endless screensaver of self-assembling quantum cryostats, made with Gemini 4 Carbon — hapticdata · 2026-11-12
- ADM 3 Cloud ranks #47 of 80 at photo enhancement and restoration — ArtificialAnlys · 2026-11-12
- Apple's ADM 3 Cloud tested: #42 at image editing, #85 at text-to-image — ArtificialAnlys · 2026-11-12
- Object-level edits are ADM 3 Cloud's strongest skill, ranking #27 of 80 — ArtificialAnlys · 2026-11-12
- Apple's ADM 3 Cloud image model ranks #42 in editing and #85 in text-to-image on Artificial Analysis arena — ArtificialAnlys · 2026-11-12
- Adobe Firefly debuts Story Builder tool for streamlined visual storytelling — LudovicCreator · 2026-11-12