Generate a model
Menu

Free AI text to 3D, No sign-up

Describe one object in a sentence. It is painted as a photo, the photo becomes a textured mesh, and you get both, in about forty seconds.

Examples:

Or drop a photo of one object

  • JPG
  • PNG
  • WebP
  • HEIC

The image is sent to a GPU for the run and not kept anywhere.

How it works

  1. Paint the photo

    one object, plain background, three-quarter view

  2. Cut it out

    background removed

  3. Predict and mesh

    shape and material together, with UVs

  4. Bake the textures

    colour, roughness and metallic

The painted photo is the whole input to the mesh stage, which is why it is shown with every result. Already have a photo? Image to 3D skips the first stage. The two mesh settings, Fast and Detailed, are compared on the image to 3D page.

Questions

How does text to 3D work?
In two stages. The description is painted as a clean studio photo of one object on a plain background, then that photo goes through the same image-to-3D engine as an upload. You get both, so if the picture is right but the mesh is not, you can build the mesh again from the same picture.
Can I see the picture before the model is built?
Yes. Tick "Check the picture first" under the description. The run stops once the picture is painted: build the model from it, or change the description and paint again until it looks right. The model follows the picture, so that is where you get the look right.
What makes a good description?
One object, named plainly, with a material and a colour: 'a worn leather armchair', 'a ceramic teapot with blue flowers'. Style words help less than you would think; the image stage already knows to make a product photo. Scenes, people and text on the object do not work. One thing at a time.
Why did I get a different model the second time?
Each run paints a new photo from a new seed, so each run is a new object. That is on purpose: generate again until you like it, or tick "Check the picture first" to settle the picture before any model is built.
Can I get an STL to print?
Yes. Every result downloads as STL too, and Repair for printing seals it into one solid that slices. Describe one chunky object with a clear silhouette, ideally on a base: 'a sitting cat figurine', 'a chess rook'.
Is it free, and who owns it?
Free, no sign-up, and yours. Both models, FLUX.2 klein for the picture and TRELLIS.2 for the mesh (Fast and Detailed are its two settings), are open weights with licences that put no claim on their output.
Do I need an account or to sign up?
No. There is no sign-up, no login and no email: open the page and use it. The AI runs on a free daily allowance per visitor, and when it is used up the page says when it comes back; the tools that run in your browser keep working in the meantime.

Next