On the Plus plan, ironing.mode=ai.auto returns ~1 MP regardless of other parameters — we measure 832×1248 from a 3605×5408 input, identical with and without removeBackground. This isn’t documented. Is there a cap, and is a higher resolution available on a higher plan as with Ghost Mannequin’s 2K/4K?
Thanks for flagging this. The Image Editing API supports input images up to 5,000 pixels on the widest side, but some AI generation features can have their own output resolution limits.
For AI Ironing specifically, I can’t confirm that 832 × 1248 is an intended fixed cap or that higher 2K or 4K output is currently available on another plan.
Given that you are consistently receiving the same resolution from a 3605 × 5408 input, including with removeBackground disabled, I’m checking with the team to confirm the expected output resolution for ironing.mode=ai.auto and whether a higher resolution option is available.
The team confirmed that AI Ironing rescales images to approximately 1 MP. Larger inputs are downscaled, which explains the 832 × 1248 output from your 3605 × 5408 input. Very small inputs are upscaled.
There’s currently no 2K or 4K output option for AI Ironing, including on a higher plan. Our ML team is working on a new model version intended to support higher output resolution, but we can’t commit to availability or a release timeline yet.