gpt-image-edit

Edit images with GPT Image 2 while preserving identity and rewriting embedded text.

31|9|Updated Apr 30, 2026
One-click install
npx skills add https://github.com/agentspace-so/runcomfy-agent-skills --skill gpt-image-edit-agentspace-so
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: gpt-image-edit
Source: https://github.com/agentspace-so/runcomfy-agent-skills/tree/main/gpt-image-edit
Command: npx skills add https://github.com/agentspace-so/runcomfy-agent-skills --skill gpt-image-edit-agentspace-so

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It solves the need to perform high-quality, identity-preserving image edits—especially edits involving embedded multilingual text—without trial-and-error prompting.

Core Features & Use Cases

  • GPT Image 2 /edit image-to-image editing: Apply targeted changes while keeping the original subject, brand mark, and framing stable.
  • Embedded text rewriting in any script: Rewrite in-image headlines/labels by quoting the exact characters (Latin, kana, CJK, Cyrillic, Arabic) and keeping layout typography consistent.
  • Multi-reference edits for controlled composition: Use up to 10 reference images with clear per-image intent to compose or blend subject identity, lighting, and scene elements.
  • Smart sibling routing guidance: Choose this model for preservation + text edits, and route to alternatives (e.g., Nano Banana Edit, Flux Kontext, text-to-image sibling) for other generation/batch needs.

Quick Start

Use the gpt-image-edit skill to replace the background and headline of your image while keeping the person and brand mark unchanged.

Frequently Asked Questions about gpt-image-edit

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I edit embedded multilingual text in an image while preserving the original layout?

To edit embedded multilingual text while preserving layout, use identity-preserving image editing via GPT Image 2. You provide the exact target characters in your prompt, and the model rewrites in-image headlines or labels while keeping typography and framing stable.

Can I use multiple reference images to compose a single edited image?

Yes, you can use multiple reference images for controlled composition. Provide up to 10 input image URLs with clear per-image intent to blend subject identity, lighting, and scene elements into the final edited output.

What is the best way to localize a headline and CTA without losing the brand mark?

The best way to localize a headline and CTA while keeping the brand mark is applying a targeted image-to-image edit. This approach rewrites text across scripts like Latin, CJK, or Arabic while keeping the original subject and framing stable.

Does GPT Image 2 editing support non-Latin scripts like CJK, Cyrillic, and Arabic?

Yes, GPT Image 2 editing supports non-Latin scripts including CJK, Cyrillic, and Arabic. You can rewrite in-image text by quoting the exact characters needed, ensuring the localized text matches the original layout typography.

What size constraints apply when sending an edit request to the image API?

When sending an edit request, the size parameter is optional and constrained to auto or supported fixed presets. You must include a prompt and an images array containing your reference URLs to satisfy execution requirements.

When should I choose identity-preserving image editing over text-to-image generation?

Choose identity-preserving image editing when you need to modify existing visuals with precise text or layout changes while keeping the subject stable. Route to text-to-image generation when you need to create entirely new visuals or perform batch generation instead.