Skip to content

How it works

From a few photos to a model of your own.

Four steps. Most of the work is ours.

  1. 01

    Choose what to train

    Start a new model and pick its type — a person, a product, a visual style or an illustrated character.

    Training a person requires that they are an adult and have given explicit consent.

  2. 02

    Upload training images

    Add a small set of clear, varied images — typically ten or more. JPEG, PNG and WebP are supported.

    Images are stored privately and used only to train your model.

  3. 03

    Train your model

    Training runs in the background and uses tokens. We’ll let you know when your model is ready.

    After successful training, your source images are deleted automatically — after 7 days by default. You can delete them sooner.

  4. 04

    Generate in AI Studio

    Select your model, describe a scene, choose a format and generate. Upscale the results you want to keep.

    Outputs are private and saved to your library.

Training set

What makes a good training set.

Do

  • Use sharp, well-lit images
  • Vary angles, backgrounds and lighting
  • Keep the subject clearly visible
  • Use only material you own or have rights to

Avoid

  • Heavy filters, blur or watermarks
  • Group photos with several people
  • Near-identical shots
  • Anyone who hasn’t consented — or anyone under 18

Tokens

One currency for everything.

Training, generation and upscaling each use tokens. Plans include a monthly allowance; packages add more whenever you need them.

See pricing

Build your first model.

Create an account in a minute. Upload when you’re ready.

Get started