AI Image Edit Analysis
ai-image-edit-analysis
Controls whether a lightweight prepass vision analysis runs before the main AI image edit. When enabled, a vision call examines the image and determines whether the requested edit would meaningfully improve it. If the prepass returns should_edit=false, the original image is served unmodified and a skip_reason is logged — no full edit credits are consumed.
| Value | Behavior |
|---|---|
true | Runs the prepass gate before processing. Costs 0.5 credits regardless of outcome. |
false | Skips the prepass gate and always runs the full edit. |
The analysis gate is on by default for packaging and off by default for all other ai-image-edit values.
When the Gate Skips an Edit
Common reasons the prepass returns should_edit=false and serves the original image unmodified:
- No packaging visible in the frame (
packaging) - The scene is not a room (
redecorate-*) - No sky visible in frame (
sunny-sky) - Pool is drained or not water-filled (
pool-clarify) - Lawn area is mulch, gravel, or hardscape rather than grass (
green-lawn) - Wrinkles appear to be intentional design elements (
wrinkle-removal-fabric) - The scene already reads as the target state (e.g. sky is already sunny, room is already staged in the requested style)
Questions
Please reach out to your Account Manager or to Support if you have any questions or feedback.