English

Scene from a photo

Give ChatGPT, Claude, Codex or another AI a photo, and it builds a scene in Blender to match. The photo sits behind the scene camera as a reference; the AI sets up the camera, objects, materials and lighting, then saves one image with the photo on the left and the render on the right. You see at once what already matches and what still needs work.

How it works

Share the photo: as a chat attachment if your AI can pass files to its tools, as a direct public link to the image, or as a file in your project folder. PNG and JPEG both work.
The AI puts the photo behind the camera and builds the scene to match: proportions, object placement, camera angle, lighting.
Render and compare. The AI renders a frame and saves a side-by-side image of the photo and the render to your project folder. Use a photo and render of up to 8 megapixels each; they are scaled down side by side, with proportions kept.
Refine. Look at the comparison and say what to change: "lower table", "wider window", "cooler light".

What you can build from a photo

An interior or a room — to visualize a renovation or new furniture.
A product — to get studio shots from any angle.
A building or a street — as the base for a scene or a game level.
A shot from a film or an ad — to recreate its composition and lighting.

Get the best result

Use a sharp photo that shows the whole subject. The less is hidden, the more accurate the scene.
Give sizes if you know them: "the room is four by five meters, the ceiling is 2.7 meters high".
Work in steps: geometry and camera first, then materials, then lighting.
Free textures can come from Poly Haven and ambientCG, and HDRI skies and models from Poly Haven — see Materials and free assets.

If a version doesn't work for you, return the scene to the checkpoint you need and ask again.

Next step: connect ChatGPT, Claude or Codex to Blender.

Try Blender Bridge

Next: Animation in Blender with AI: keyframes, camera orbits, rigging, physics