Simulations are illustrative and are not a medical prediction or medical advice. Individual results vary — consult a board-certified plastic surgeon.
⚡ GenStudio
📷 Tap to add a photo
Typography edit — change words on signs, labels, posters, screenshots and documents. Say exactly what to replace, in quotes: change "SALE" to "SOLD". The same font, size, weight, colour and layout are kept; with Lock everything else on, every pixel outside the changed text is your original file — nothing else can drift. Big images (phone screenshots, scans) automatically get a second pass that redraws the changed words at full resolution — about twice the time.
Standard's clean full-image edit (keeps the background, removes the shirt) plus depth + edge guidance to fix chest size and areola placement — nothing else changed. If the shirt won't fully come off, lower Edges; if the chest shape drifts, raise Depth.
Clothing edge reach — how far past the old garment to repaint (fixes leftover fabric on skin):
Clothes only (default): a clothing segmenter finds your garments and only those pixels are repainted — skin, face, hair, pose and background stay bit-identical. Body: whole body repainted inside its own outline (skin included), face and background untouched. Keep face: face untouched, everything else editable (change the scene). Whole image: everything editable. A depth map of you guides all of them; Edges holds contours visible through tight fabric. "Face detail" lightly refines skin texture (not used with Clothes only). About 2–3× slower than Standard.
🦴 Add a pose reference (optional)
Optional: a photo of anyone in the pose you want. Only their skeleton is used — your face and identity come from your photo.
🧍 Add a body reference (optional)
Optional: a real photo of the same person showing the body area you're changing (e.g. shirtless at the pool). The model copies their actual build, skin and markings instead of guessing. In Experimental it replaces the edge map.
⚠️ The engine is in Images-only mode (small GPU) — video may be slow or run out of memory. Toggle it off and choose "Images + video" for the fast GPU.
First generation of a session takes ~1–2 min (model loads into VRAM); after that it's quick. Engine stays on until you toggle it off.
No generations yet. Create one!
Qwen is off — tap the Qwen button in the header to start it (separate GPU pod, billed only while on).
For IDE tools (Continue / Cline / aider) while it's on: OpenAI-compatible endpoint on this Mac — shown here when running.
Tap an item to view it, restore it to the Library, or delete it forever. Deleted results, originals and prompts stay here for 60 days, then disappear on their own.
Engine auto-stops after 20 min without generations.
Prompt instructions (style prefix/suffix, negatives, pose, body and face-pass lines) live in runpod-genstudio/prompts.toml on the Mac and are re-read before every generation.
Recording
0.00 s
Drag the box to move it, the corners to resize
Guided capture
Guided capture
3
0:00
Video ready for capture and clipping
Rhinoplasty
Edit optionsLocked
Each position is generated with a fixed preset built from your options — nothing to type. Results appear one by one; press and hold a result to see your original underneath.
Results
Loading…
Tap a finished position to view it; press and hold to see your original. Use full screen if you want a screenshot — results can't be downloaded here. You can go back and stage another procedure while these finish.