gpt-image-2
Advanced image generation and editing with reasoning capabilities, editing tools and platform-specific settings. Use when creating or editing images with an AI image model.
SKILL.mdpsk/data/skills/gpt-image-2/SKILL.md
GitHubGPT Image 2 (AI Image Generation)
Produce and edit images with an image model using structured, platform-aware prompts.
Generation workflow
- Intent — subject, purpose and platform (IG, YouTube, blog header...).
- Prompt structure — write prompts as:
[subject] + [action/pose] + [environment] + [composition/camera] + [lighting] + [style/medium] + [mood] + [negative constraints]. - Reasoning step — before prompting, reason about what visually communicates the message (symbols, setting, expression) instead of literal depiction.
- Platform settings — set aspect/size per target: IG post 1080x1080, reel/story 1080x1920, YouTube thumb 1280x720, X header 1500x500, Dribbble 400x300.
- Iterate — refine one parameter at a time; keep a prompt log.
Editing workflow
- Addition / removal / replacement: describe precisely what changes and what must stay identical (lighting, perspective, grain).
- Inpainting: specify the region; outpainting: specify extension direction and content continuity.
Quality rules
- One clear subject; specify camera (e.g. "85mm portrait", "isometric", "top-down flat lay").
- Always declare text that must appear in-image, in quotes, short (models garble long text).
- Check hands, text, reflections and logos in every output before delivering.
Output
Final image(s) + the exact prompt used + suggested variations.