Upload your photo
Drop a photo or click to choose. Indoor scenes work best.
Depth map
Generated by Depth Anything V2. Used as ControlNet condition for the next step.
Photo
Depth
🔑 Use your own Anthropic API key (optional)
Stored only in your browser (localStorage). Sent with each Enhance
request to api.anthropic.com via this server; never
logged. Get a key at
console.anthropic.com.
Pick up to 3 styles
Click cards to select. Each style will be rendered separately in the next step so you can compare and pick the best one.
0 / 3 selected
Generate
Adjust knobs, then press Render to render —.
Original
Stylized
Not yet rendered.
0.75
Balanced — visible style change while keeping structure (recommended)
Off = stronger restyle (recommended). On = preserves edges but flattens differences between styles.
After you pick a style, it auto-renders at 1536 px before the click step.
Pick the version to use
Click a thumbnail below to select. That image becomes the working scene for click + interact.
Click an object
Pick what to do with the next click — lift to 3D, erase + auto-fill, or replace via prompt.
No edits to undo yet.
1 Type what to put there
2 Click the object on the image
3 Press Generate
Type a prompt above, click the object on the image to mark it, then press Generate.
🤖
Auto-detected objects
—
Pick one to lift, or click directly on the image below for manual selection.
Detecting objects with SAM 2 + CLIP… 0s
click select object ·
⇧ shift+click add point ·
⇧+drag brush add ·
⌥ alt+click sub point ·
⌥+drag brush erase
36 px
Preview
part
scene
g=0.0
—
Click an object on the image.
Click an object to begin.
Stretch
Drag to fix perspective foreshortening — e.g. tilted curtain comes out short, push Y up.
TripoSR multi-view debug renders — model's native output, before mesh / splat conversion.
If these look like the source object, the model worked; if they look gray and shapeless,
the model itself produced nothing usable for this input.