Subject + product → one photo

A person and a product, photographed together in a shot that never happened.

What it takes

• A picture of the subject
• A picture of the object
• A line describing the scene they share

What it gives back

• One photoreal frame — and the best possible first frame for a video

What actually runs

• Qwen-Image-Edit 2509 in fp8, with the 8-step Lightning LoRA.
• Both pictures are tokenised as separate references, so the prompt can address Picture 1 and Picture 2 and both survive into the result.
• Feed the result straight into image-to-video: frame one is then locked by construction, which is the single biggest lever against a likeness drifting mid-clip.

Worth knowing. Compose first, animate second. A bad frame costs seconds to redo; a bad clip costs minutes.

Running it

The .json attached to this listing is the graph itself — the one the Forge runs, not a tidied-up copy. Values in {{double braces}} are the placeholders filled in per job; replace them with real values and it runs in any ComfyUI new enough for the nodes it uses.

Full setup — the model manifest, the directories, the traps — is at /workflows/compose.

Licence and credit

The graph is ours and you may use it for anything, including commercially. The models it loads are not ours: each carries its own licence from whoever trained it, and that is the one that governs what you may do with the output.