Docs menu

Creating with AI

Generate an image

One screen, three modes: describe it and let the Director build it, take full manual control in Advanced, or switch to video without losing your setup.

Three modes

A pill at the top of the generate screen switches between them. Director is the default and is covered on its own page. Advanced, described below, exposes every control directly. Video swaps in the video panel inline, without leaving the screen.

Creative direction

Five cards let you set the ingredients of a shot before you write a word: Product, Style, Shot type, Environment, and Character. Each opens a picker.

Shot typesStylesEnvironments
Product HeroMinimalStudio White
LookbookLuxuryStudio Black
LifestyleEditorialNatural Light
FlatlayBoldGolden Hour
Street StyleSoftUrban / Nature / Interior / Abstract
CampaignDark

The Environment picker also has a "Create your own" option: name a scene and add an optional description, and it's saved for this brand alongside the built-in ones.

Camera settings

A camera treatment is never applied silently. Camera defaults to None; choose one only when you want it. When you do, four dropdowns compose the shot:

  • Camera: Cinema 70mm Film, Lo-Fi Toy Camera, Camcorder, 8mm Film, 35mm Film, DSLR, Medium Format, Polaroid
  • Lens: Anamorphic Classic, Prime Lens Classic, Anamorphic, Tilt-Shift, Macro, Fisheye, Telephoto, or Standard
  • Focal length: 14mm to 200mm
  • Aperture: f/1.4 to f/22

A separate Platform setting (E-commerce, Social, Ad, Editorial) nudges composition toward where the image will actually run.

Reference images, output, and models

Attach up to six reference images, reorder them, and remove any at any time. Choose an aspect ratio (1:1, 4:5, 3:4, 9:16, 16:9, 4:3, or Auto, which reads your prompt for words like "vertical" or "widescreen"), a resolution up to 8K, an output format, and how many outputs to generate at once, up to six.

The model picker offers four ways to choose:

  • Auto: the AI picks what's best for the shot you've described.
  • Cinematic: tuned for professional, filmic output.
  • Multiple: run up to four models on the same prompt at once.
  • All models: pick a specific model yourself from the full list, including Flux Schnell for near-instant drafts and GPT Image 2, Imagen 4, or Nano Banana Pro for finished quality.

Prompt help and saved recipes

Enhance rewrites your prompt in place. The expanded editor's chat assist can also add detail or suggest variations. If a combination of product, style, shot, environment, and character is working well, save it as a recipe (up to twenty per brand) and reload it with one click next time.

A couple of buttons aren't wired up yet

"Image to prompt" and the chat attachment icon both show a "coming soon" message. "Convert to JSON" doesn't do anything yet either.

What happens after you click Generate

Every result saves to your Library automatically with no status set, ready for you to approve, reject, or leave as-is. If a generation fails, you'll see the error directly on the tile with a way to dismiss it, and any tokens spent on that failed attempt are refunded.

See also