Wan 2.7 image-to-image from Alibaba merges or edits one to nine reference images into a new image under a single instruction. The prompt says what changes and what must stay, so garments, products or faces survive the edit; this endpoint is uncensored.
Filtering differs by endpoint: Image editing — Uncensored, Text to image — Softened.
Workbench
These examples were written for this endpoint. Replace the prompt and sample media URLs with your own accessible inputs, then check the current quote and model availability before submitting.
Outputwardrobe-swap Two labelled references: one supplies the person, the other supplies the garment.
Outputfour-ref-album-cover Four references under one instruction - subject, side light, prop and atmosphere, each named to its source - which sits in the middle of what `image_urls` will take. Worth running before you try nine: at this size the failure mode is a clause being dropped rather than the whole merge collapsing, and with four sources it is still possible to tell which one went missing. Deep red side light on a sweat-soaked back keeps that check visual.
- 2k
- 3:4
Outputstudio-to-frozen-lake Face, proportions and pose are frozen; the white studio becomes a cracked-ice lake at blue hour, the outfit becomes winter knit, and the whole image is relit cold with a weak sun grazing one shoulder. The instruction then desaturates everything except one scarlet scarf — a single colour left alive is what keeps a heavy restyle from looking like a filter.
This example starts at $0.03 per image. The Playground updates the estimate when you change the settings.