How to Use the ArtCraft AI Image Generator

ArtCraft can do everything a typical AI image generator does: type a prompt, get an image. Where it differs is everything that happens before and after the prompt. You can compose layers, stage a 3D scene, pose characters, and paint masks, so the model renders what you planned instead of what it guessed.

The five image workflows

Workflow 1: Text to image

1. Open the Image tool

   In the web app or the desktop app, choose Image from the start screen.

2. Pick a model

   Start with an inexpensive model like Nano Banana or GPT-Image-1.5 while you explore. Switch to Nano Banana Pro, GPT Image 2, FLUX 1.1 pro ultra, or Seedream 5.0 Pro when you are ready for final quality.

3. Write a structured prompt

   Describe the subject, setting, composition, lighting, and style. Use the template below.

4. Iterate one change at a time

   Keep what works and change a single variable per generation, such as lighting, angle, or style, so you learn what each word does.

A prompt structure that works across models:

[subject + key details], [action or pose], in [setting],
[shot type and camera angle], [lighting and mood], [visual style or medium]

Example:

an elderly clockmaker with round glasses, inspecting a pocket watch,
in a cluttered workshop full of brass gears, close-up shot at eye level,
warm window light with dust in the air, 35mm film photograph

Workflow 2: Prompted image editing

Use an editing-capable model such as Nano Banana Pro or GPT Image and describe the change in plain language. Spell out what should stay the same:

Keep the character, pose, and framing exactly the same.
Change the background to a rainy neon street at night.
Match the lighting on the character to the new scene.

Workflow 3: Canvas editing with masks

When a prompt edit changes too much, switch to canvas editing. Paint a mask over only the area you want to change, describe the new content, and inpaint. Everything outside the mask stays untouched, which makes it ideal for fixing hands, replacing a prop, or cleaning up text.

Workflow 4: 2D compositing

Build an image from parts:

1. Import your source images and use background removal to cut out subjects.
2. Arrange the cutouts as layers on the canvas and adjust their size and position.
3. Sketch rough shapes with the drawing tools to show where things should go.
4. Render the composition with an image model so lighting and style blend together.

This is the fastest way to get a specific layout, such as a product shot, poster, or thumbnail, without fighting a prompt.

Workflow 5: 3D compositing and posing

For the most control, stage the shot in 3D:

• Use image to 3D mesh to turn props or characters into objects you can rotate.
• Place backdrops, foreground elements, and props at real depths.
• Pose characters with a mannequin and use character identity transfer to keep a consistent character.
• Use Image to Location to keep one environment consistent across several shots.
• Set the camera angle and distance, then render.

Because the geometry is fixed before generation, you can produce several consistent shots of the same scene, which is very hard with prompts alone.

Which image model should you use?

Midjourney is also available if you connect your own Midjourney subscription. The best model depends on your subject, so run the same prompt through two or three models before committing to one.

Common problems and fixes

• The model ignores your layout. Stop describing the layout in words and block it out with 2D layers or a 3D scene instead.
• Characters change between images. Use character posing with identity transfer, and keep the same reference image across generations.
• Edits change too much. Switch from prompted editing to a masked canvas edit.
• A generation fails. Try a different model. The README marks some models as temporarily disabled, so availability changes over time.

Next, learn how to turn your images into motion with the ArtCraft video generator guide.