Reposition a subject — head turn, arm placement, posture — without a reshoot, keeping the same face and clothing.
Pose changes are the hardest class of photo edit because the model has to invent anatomy that was never photographed — the far side of a shoulder, a hand that was behind a hip — while keeping the person recognisably the same. Instruction-tuned image models handle this far better than warp-and-clone editing ever did.
Describe the change in one sentence and the editor rebuilds the affected area with matching light and perspective. Small, physical instructions work best: 'turn the head slightly toward the camera', 'lower the raised arm to the side'.
Upload the shot: A photo where the subject is unobstructed gives the model the most to work with.
Ask for one change at a time: 'Straighten the posture and square the shoulders' beats a paragraph of instructions. Stack edits by re-running on the result.
Continue editing or download: Every result becomes the new input, so you can refine a pose across several passes.
Photo-consistent anatomy: Reconstructed limbs inherit the original lighting, skin tone and lens perspective.
Chain edits: Continue Editing feeds the result back in, so pose, outfit and background changes can be layered.
No rig, no 3D: Nothing to model or pose manually — it is a sentence of plain English.
How accurate are AI pose changes? Small adjustments — head turns, arm placement, posture — are reliable. Full body rotations beyond about 45 degrees start to invent detail the photo never contained.
Does it keep the same clothing? Yes, unless your instruction asks otherwise. The garment is re-draped to fit the new pose.
Is it free? Signed-in accounts get 10 free credits every day, and new users can try Pro free for 24 hours — cancel anytime.