What Is ArchiGPT?
ArchiGPT is ArchiVinci's AI image editor. You give it a picture and an instruction in plain words, the way you would brief a colleague: "replace the sofa with a green velvet one", "change the brick to white stone", "add a timber pergola over the terrace". It returns a new version of your picture with that change made.
In the app, ArchiGPT is the Edit module. It works on renders from any 3D or rendering software, on photos of real rooms and buildings, and on images you made in ArchiVinci's other modules. It is an editor, not a chat assistant: there is no conversation, only a picture, a request and a result. A new account gets 3 free credits.
How Does ArchiGPT Work?
Open the Edit module and upload one image: JPG, PNG or WebP, up to 5 MB.
Write what should change in the Prompt field. In this module the prompt is required.
If it helps, pick "Add object" tags such as Sofa, Plants, People or Window, or upload one picture of an object you want in the scene.
Press Generate. The result opens in a before/after viewer next to your original.
Download the result, or keep editing it.
The example below is one run in the Edit module on the default engine, with this request:
Replace the sofa with a green velvet sofa and add a floor lamp next to it
The editor is asked to keep everything else in the image, the composition and the camera angle. It is an instruction to an AI model, not a lock, so check small details.
Edit One Area and Keep the Rest
When only one part of the picture should change, mark it. Under the image there are three area tools: Select paints the area with a brush, Select area outlines it with a lasso, and Remove takes parts back out of what you marked. Press Done, write the request and generate.
A marked area can be changed in three ways:
With words: "turn this wall into exposed concrete".
With a material: Apply texture offers a library of 103 materials in 15 categories, or your own sample image.
With an object: upload a picture of the product and ask for it in the marked spot.
The result of an area edit becomes the next picture to edit, so you can work through a scene one change at a time: the floor, then the sofa, then the lamp.
What You Can Do with ArchiGPT
ArchiGPT is one editor behind several jobs. Each of them has its own page with examples and prompts:
Job | What you do | Read more |
|---|---|---|
Restyle part of a room | Ask for new furniture, colors or finishes | |
Update a facade | Ask for new cladding, windows or a roof | |
Empty a room | Mark the furniture and ask to remove it | |
Add a product | Upload its picture and say where it goes | |
Change the light | Describe the time of day or the mood | |
Change one surface | Mark it and apply a texture |
ArchiGPT Engines and Cost
The Render engine menu decides which AI model makes the edit:
ArchiVinci 4.0 is the default. One image costs one credit, and with a plan this engine is unlimited.
Nano Banana 2, GPT Image 2, Seedream 5.0 Lite and Qwen Image 2 are partner models. They are paid with coins and make one image per run.
With a plan you can also ask for up to 4 versions of an edit in one run; a free account makes one image per run. Plans and coin packages are on the Pricing page.
What ArchiGPT Gives You Back
Every result is a JPEG image in the proportions of your upload, without a watermark. With a plan you can upscale a result up to 4096 px on the long side, and Animate turns it into a 10-second video that is paid with coins. Your results stay in the render history for 7 days, so download the ones you want to keep.
ArchiGPT edits pictures. It does not produce CAD files, measured drawings or 3D models, and it has no batch mode: one image is edited at a time.
Frequently asked questions about ArchiGPT
Is ArchiGPT a chatbot, or built on ChatGPT?
No. ArchiGPT is ArchiVinci's own image editor: you upload a picture and write one instruction, and it returns the edited picture. There is no chat. Its default engine is ArchiVinci 4.0; GPT Image 2 is one of four partner models you can choose in the Render engine menu.
How is ArchiGPT different from the Interior and Exterior modules?
Interior and Exterior render a whole image again from a sketch, a 3D view or a photo, in the look you describe. ArchiGPT changes what you ask for and is asked to keep the rest, and it can limit the change to an area you mark. Use Interior or Exterior to create the render and ArchiGPT to revise it.
Can ArchiGPT edit a render made in another program?
Yes. ArchiGPT takes any image as a JPG, PNG or WebP file up to 5 MB: a render exported from SketchUp, Revit, Blender, 3ds Max or Lumion, a photo, or a result from another ArchiVinci module.
Does the rest of my image stay exactly the same in ArchiGPT?
In an edit of a marked area, that is how it is built: only the marked area and a narrow margin around it are redrawn. In an edit of the whole picture the model redraws the image and is asked to keep everything you did not mention, so compare the result with your original.
Can I go back to an earlier version in ArchiGPT?
Yes. The render history lists your recent results for 7 days, and Re-use makes any of them the next picture to edit. Your uploaded original is never changed.