stepfun-img-skill

StepFun image generation and editing skill for Codex, built on Step Plan and step-image-edit-2.

What this repo does

• Text-to-image generation with StepFun Step Plan
• Image editing with StepFun Step Plan
• OpenAI SDK based calling flow
• Codex-friendly scripts and docs

Quick Start

1. Create ~/.stepfun-img/secret.json:

{ "apiKey": "YOUR_STEP_API_KEY" }

2. Generate an image:

python scripts/generate_image.py "your prompt here" --model step-image-edit-2 --size 1024x1024 --out outputs/image.png

3. Edit an image:

python scripts/edit_image.py "path/to/source.png" "your edit prompt" --model step-image-edit-2 --out outputs/edited_image.png

Parameters

• model: step-image-edit-2 is recommended under Step Plan. step-2x-large and step-1x-medium are available on /v1, not Step Plan.
• prompt: text description, up to 512 characters.
• size: image size. For step-image-edit-2, use 1024x1024, 768x1360, 896x1184, 1360x768, or 1184x896.
• seed: random seed for more repeatable results.
• steps: generation steps. Higher values usually refine more slowly but can add detail.
• cfg_scale: guidance strength, meaning how strongly the model follows the prompt.
• negative_prompt: unwanted features to suppress. Only step-image-edit-2 supports it.
• text_mode: optimize for text-heavy scenes. Only step-image-edit-2 supports it.
• style_reference: style reference for step-1x-medium.

Chinese search keywords

StepFun 图片生成, StepFun 图像编辑, StepFun 文生图, StepFun AI 作图, Step Plan 图像接口, step-image-edit-2, OpenAI SDK 图片生成, Codex skill, 中文提示词, 图像生成, 图像编辑, AI 绘图, 生成图片, 编辑图片

Files

• SKILL.md: skill instructions
• scripts/generate_image.py: text-to-image runner
• scripts/edit_image.py: image edit runner
• references/api_spec.md: API notes

Notes

• Base URL: https://api.stepfun.com/step_plan/v1
• Supported model: step-image-edit-2
• The API returns url or b64_json