Skip to content
0/3000
Cost 10 credits

My Library

AI image for 3D

Most image to 3D problems start with the image. Generate a clean, single-subject reference image from a prompt first — with Nano Banana 2 or Seedream 5 Lite — then send it to image to 3D.

10 credits an image
Nano Banana 2 or Seedream 5 Lite
One subject on a plain background
Opens in image to 3D

What makes a good AI image for 3D

An image to 3D model reads the silhouette, the lighting and the background. A reference built for it avoids the usual failures.

One subject, whole in frame

Nothing cropped at the edges and nothing else in the picture.

Plain background

A flat light-gray or white background keeps the outline clean.

Even, soft light

Strong shadows can bake into the texture, as Meshy's image to 3D guide warns; soft light keeps the colors true.

A three-quarter view

An angle that shows the front and one side gives the model more to work from than a flat front view.

Reference images that went on to image to 3D

Each input in the corner was generated with Nano Banana 2, then turned into the model on the poster. Open a card to orbit the result.

Uploaded picture: Chibi knight

Chibi knight

Tripo H3.1

Full figure, plain gray backdrop; rebuilt with Tripo H3.1.

Uploaded picture: Treasure chest

Treasure chest

Hunyuan3D 3.1 Rapid

Three-quarter view of a single prop; rebuilt with Hunyuan3D 3.1 Rapid.

Uploaded picture: Sci-fi helmet

Sci-fi helmet

Rodin Gen-2.5

Soft even light keeps the visor readable; rebuilt with Rodin Gen-2.5.

Uploaded picture: Dragon figurine

Dragon figurine

Hitem3D 3.0

One subject on its own base; rebuilt with Hitem3D 3.0.

Uploaded picture: Succulent pot

Succulent pot

TRELLIS.2

Nothing cropped at the edges; rebuilt with TRELLIS.2.

Uploaded picture: Toy robot

Toy robot

Tripo H3.1

Whole body in frame, no props around it; rebuilt with Tripo H3.1.

AI image for 3D at a glance

AI image for 3D in numbers

2 image models

Nano Banana 2 and Seedream 5 Lite.

10 credits

Per image on either Nano Banana 2 or Seedream 5 Lite at the default settings.

1 subject

The image to 3D models reconstruct one object best.

2 steps

Generate the image here, then run image to 3D on it.

From prompt to model in three steps

The reference image and the 3D model are two separate runs: 10 credits for the picture, then the image to 3D price of the model you pick.

  1. 1

    Name the subject and its material, then add "full body in frame, plain light-gray background, soft even lighting". The default 1:1 ratio at 1K leaves room around a centred subject.

    The AI image for 3D form on Text to Image with a knight prompt typed in, 1:1 aspect ratio and 1K resolution selected, and a cost line of 10 credits beside the Nano Banana 2 knight sample
  2. 2

    Both cost 10 credits at the default settings. Switch to Image to Image when you already have a photo and only need the background cleaned or the subject isolated.

  3. 3

    Image to 3D opens with the picture already in the form; pick a 3D model there. The default Tripo H3.1 costs 38 credits with a standard texture; Hunyuan3D 3.1 Rapid starts at 29.

Limits of the reference-image generator

A picture is still one view.

Hidden sides are still guessed

The 3D step infers the back; generate extra views for multi-view to 3D if it matters.

Two charges

The reference image costs 10 credits on either model, and the 3D model is billed as a separate run, from 29 credits on Hunyuan3D 3.1 Rapid or 38 on the default Tripo H3.1.

Text in images is unreliable

Lettering on a generated object rarely survives into 3D.

Questions about the reference-image generator

Answers about preparing reference images.








Start with a clean image

Generate a reference image, then turn it into a 3D model.