Step Image

Step Image

Generate and edit images using StepFun's step-image-edit-2 model.

When to use

• Create an image from a text prompt
• Edit an existing image with a text instruction
• Generate or revise visual assets for a project
• Apply style transfer, background replacement, or object changes

When not to use

• SVG/vector icons, logos, or illustrations that belong in code
• Simple shapes, diagrams, or wireframes better built in SVG/HTML/CSS
• Tasks where the user clearly wants deterministic code-native output

Workflow

1. Collect the prompt, exact text, input images, and constraints from the user.
2. Label each input image by role: edit target, style reference, supporting insert.
3. If the edit target is a local file, inspect it before editing.
4. Run scripts/image_gen.py with the right command and arguments.
5. Check the output for subject, style, text accuracy, and invariants.
6. Iterate with one targeted change if the result needs adjustment.

Prompt rules

Write prompts with enough detail for image models:

• Subject and scene
• Composition and camera or layout
• Style, mood, lighting, and color palette
• Text that must appear in the image, quoted exactly
• Constraints such as "keep the same character", "preserve the logo", or "do not change the background"

For edits, state what changes and what stays unchanged. Repeat invariants on every iteration.

Supported sizes

• 1024x1024
• 768x1360
• 896x1184
• 1360x768
• 1184x896

Environment

• Set STEP_API_KEY before running CLI commands.
• Base URL: https://api.stepfun.com/step_plan/v1
• Default model: step-image-edit-2
• Optional model: step-1x-medium for style-reference image edits

Output

• The CLI writes final images under output/step_image/ by default.
• Use --out to choose a specific path.
• Do not overwrite existing assets unless the user asked for replacement.

Reference map

• references/cli.md: CLI usage and recipes
• references/image-api.md: API parameters and limits
• references/prompting.md: prompt structure and iteration guidance
• references/sample-prompts.md: prompt recipes by use case