Qwen-Image 2.1 Glamour Portrait — 4-Mode Auto-Prompt Workflow (SFW, 16GB) - v1.3 Showcase

12 images picked from one batch of 24 (4 outfit modes x 6 seeds, seeds 888201-888206). Each image is a single generation at 1792x2240: no upscaler, no face fix, no inpainting, no editing. Settings: Viggle Turbo v0.3 (6 steps, euler) on the Qwen-Image 2.1 graph, with Qwen21 Glamour Realism LoRA v1.0 at strength 0.6. About 23 seconds per image on an RTX 5080 (the standard 25-step graph takes about 70 seconds). Drag any PNG into ComfyUI to get the workflow it was made with. Note: the embedded workflow is the LoRA-on test variant. The downloadable workflow has the LoRA node bypassed by default; enable it and pick the LoRA file yourself. What we checked: 24 images, 1 rejected by our own check (face not matching the intended look), no anatomy problems found in the other 23. Tested at LoRA strength 0.6 only, one image per seed/mode, so treat this as a sample, not a guarantee. Requirements: the Viggle Turbo model file and its ViggleTurboSigmas node (not included in the workflow file; links are in the model description).