Impossible-Studio-TWImpossible-Studio-TW/claude-gpt-image2-mcp/skills/gpt-image-edit/SKILL.md

gpt-image-edit

Skill0 starsChanged 7 months agoMITView on GitHub

Edit or transform images with OpenAI's gpt-image-2 / gpt-image-1, with one or more reference images. Triggers when the user types "/gpt-image-edit", attaches an image and asks to modify it, says "改這張圖", "把這張圖換成 X", "用這張圖當參考生 X", "多圖參考", or wants character-lock / multi-reference style transfer. gpt-image-2 multi-image is true reference fusion; gpt-image-1 multi-image is weak fusion only — useful for style cues, not strict identity preservation.

4 other projects use a version of this.

Adopted by

Most recently taken up by

genmedia-labs/skillstheir own version, first seen Aug 12, 2026
prime-skills/runcomfy-agent-skillstheir own version, first seen May 15, 2026
runcomfy-com/skillstheir own version, first seen May 15, 2026
agentspace-so/runcomfy-skillstheir own version, first seen Apr 29, 2026

Other versions4 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 4
prime-skills/runcomfy-agent-skills★ 41same name
genmedia-labs/skills★ 26same name
runcomfy-com/skills★ 16same name
agentspace-so/runcomfy-skills★ 4same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repositoryOne other file it gives its agents.
- gpt-imageSkill

The file

We split this skill into the rules it gives an agent

2 rules. 0 are common, 0 are distinctive, and 2 are very specific to this project.

0
Common
Most of the 4 edited versions we read keep these.
0
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.

Common
Distinctive
Very specific

Very specific:
When the user wants to edit an existing image OR generate a new image using one or more references, call mcp__gpt-image2__gpt_image_edit.
line 8
Very specific:
If user doesn't specify model/quality/size, ask before calling:
line 44

“Common” is measured: kept by at least 6 in 10 of the 4 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.

Files

SKILL.md

/gpt-image-edit — edit / transform with reference images

When the user wants to edit an existing image OR generate a new image using one or more references, call mcp__gpt-image2__gpt_image_edit.

Two main modes

Single-reference edit — pass one image, the tool transforms it based on the prompt:

"把這張海報的文字改成 X" → single image + text-change prompt
"把這張照片的背景換成沙灘" → single image + background-change prompt
With optional mask_path: only edit the transparent areas of the mask

Multi-reference (1-5 images) — pass several images for character/style fusion:

"餵這兩張參考圖，生女主角在咖啡廳的生活照" → list of images + scene prompt
"用這 3 張圖鎖定角色，生穿不同衣服的版本" → character lock attempt
gpt-image-2 handles this well; gpt-image-1 is fusion-only (style/composition cues, not strict identity)

Invocation patterns

"改 [path]，加上 X" → single image edit
"用這兩張參考 [path1] [path2] 生 [scene]" → multi-image
"/gpt-image-edit prompt: X images: [paths]" → explicit

Parameters

Arg	Default	Notes
`prompt`	(required)	English works best; specify scene + lighting + mood
`image_paths`	(required)	List of 1+ paths. Plugin auto-resizes images >4MB or >1024px to a temp PNG before upload
`mask_path`	None	Optional PNG mask (transparent = edit, opaque = keep). Only meaningful for single-image edits
`model`	`gpt-image-2`	gpt-image-2 (best, needs org verify) / gpt-image-1 (fallback)
`size`	`1024x1024`	gpt-image-1 supports only 1024x1024, 1024x1536, 1536x1024
`quality`	`auto`	low / medium / high / auto
`save_to`	(auto)	Absolute path

Cost-sensitive: ask before defaulting

If user doesn't specify model/quality/size, ask before calling:

要哪一組？
(A) 草稿對比 — gpt-image-1 + low + 方形 — ~$0.02-0.03
(B) 標準 — gpt-image-1 + medium + 直幅 — ~$0.05-0.08
(C) 最強參考 — gpt-image-2 + high + 直幅 — ~$0.18-0.20 (needs org verify, real character lock)

Edits cost more than pure generation because input image tokens add up.

Honesty about identity preservation

gpt-image-1 multi-image is weak fusion, not true character lock. If the user is trying to preserve a specific character's face across many scenes:

gpt-image-2 is significantly better (real reference fusion)
For best character lock, dedicated tools (Dreamina/即夢's character lock, Midjourney v7 character ref, Flux PuLID, custom LoRAs) outperform OpenAI's image edit
Set this expectation upfront if user wants character consistency

Common errors

Same as /gpt-image plus:

Error	Cause	Fix
`[Error: image not found at ...]`	Wrong path	Check spelling, use absolute path, expand `~`
`[Error preparing images: ...]`	Pillow can't read the file	File may be corrupt or unsupported format
MCP timeout (60s) on high quality	Edit + high quality takes 60-120s	Image still saves to disk — read from save location

When to use this vs `/gpt-image`

Pure new generation (no reference) → use /gpt-image
Edit an existing image OR generate with reference images → use /gpt-image-edit
Character consistency across scenes → try this with multiple references; if weak, recommend dedicated character-lock tools

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

It worked
Partly
It didn't

What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free