Put your own character into any frame - a photo, a film still, a concept painting, an ad poster, a cosplay mock-up - with one Qwen-Image 2.1 edit pass. The frame is the canvas: the pose, the clothes, the other people, the camera, the light and the place stay; only the woman becomes the character on your reference picture. A depth map of the frame holds the pose and the sizes.
What it does
Input: a scene frame, a picture of your character (a face portrait is enough), and four short text boxes.
Output: the same frame with your character in it, at the frame's own pixel size.
One pass, 6 steps (Qwen-Image 2.1 int8 + the Viggle turbo LoRA), no CFG.
Why it is more than a graph
Classify first. A frame where the woman's face is not visible (back view, hidden face, head out of frame) must not get a face swap - it turns her toward the camera. The workflow has a separate hair-only mode for these frames.
One pass. A second full pass over the result degrades style and anatomy, so the graph is built for exactly one.
Keep rules. The instruction says what must stay exactly as in the frame - her expression, the clothes, what she holds, the light and the grain - so the swap does not tidy the picture up.
The prompt writes itself. Two questions for any vision model give you the pose sentence and the list of details to keep; the kit has them word for word.
Two modes, one link to change
A · face visible - the swap (default wiring)
B · no face - hair only, no face is ever drawn
Good for
photos and film stills: same light, same lens, same grain;
concept art and storyboards (one line turns a drawn frame into a photo);
ad posters and product mock-ups: the product stays in her hands;
cosplay mock-ups: the costume of the frame stays on your character.
Measured on the studio's farm (the showcase pictures, 832×1216 frames, RTX 3080 Ti / 3070 Ti boxes under normal load)
mode A swap: 41-65 s per picture, plus 11-12 s for the depth map; mode B hair-only: 23 s;
large frames (1396×2496, 1632×2880) on RTX 3080 Ti: 49-51 s for the full swap, depth map included.
Honest limits
One woman per frame in this version; with several women the swap may touch the wrong one.
The model works at about one megapixel; larger frames are resized back and come out softer than the source.
You fill the pose sentence and the details list yourself (two questions for any vision model are included). The studio's automatic classifier is not part of the workflow.
Hair colour is what drifts first: say it in the identity sentence and render once more when it is wrong.
Small faces (a few dozen pixels) take less of the plate's detail.
Text and logos in the frame are removed; add a poster's lettering afterwards.