Yummy Gen Image
/install yummy-gen-image
Generate Image
Create or edit images with yummycli gemini nanobanana.
When to Use
Load this skill when the user asks to generate, create, or edit an image using AI — including text-to-image generation, style transfer, image editing with a reference photo, or multi-image compositing.
Prerequisite: Apply the
yummy-sharedskill first.
This skill uses one command for all Gemini image flows:
- Prompt-only generation
- Single-image editing
- Multi-image reference editing
Command Contract
Two equivalent entry points are available:
| Entry point | When to use |
|---|---|
yummycli gemini nanobanana |
Default — human-friendly, Gemini presets applied |
yummycli image generate --provider gemini |
Scripting / automation — explicit, provider-agnostic form |
Both share the same flags and the same Gemini defaults. Prefer gemini nanobanana unless the task explicitly requires the provider-agnostic form.
Basic usage:
yummycli gemini nanobanana --prompt "\x3Cprompt>"
Add one or more reference images when needed:
yummycli gemini nanobanana \
--prompt "\x3Cprompt>" \
--input-image ./source-a.png \
--input-image ./source-b.png
Optional output controls:
--output \x3Cfile>
--model \x3Cmodel>
--aspect-ratio \x3Cratio>
--image-size \x3Csize>
Default values when omitted: --aspect-ratio 16:9, --image-size 1K, --model gemini-3.1-flash-image-preview.
Execution Rules
- Use prompt-only generation when no reference images are provided.
- Use one
--input-imageflag per local image file. - Preserve the user-provided order of reference images.
- Prefer the default model unless the user asks for a specific model.
Model Selection
Use the following model mapping rules:
- If the user explicitly says
gemini pro,pro model, or simplyproin the model-selection context, use:
--model gemini-3-pro-image-preview
- If the user explicitly says
gemini flash,flash model, or simplyflashin the model-selection context, use:
--model gemini-3.1-flash-image-preview
- If the user does not explicitly request a model, omit
--modeland letyummycliuse its default model.
Do not switch models implicitly from vague quality words alone. Only apply the pro or flash mapping when the user's wording clearly refers to model choice.
Model Compatibility
Aspect ratio
| Model | Supported values |
|---|---|
gemini-3.1-flash-image-preview |
1:1 1:4 1:8 2:3 3:2 3:4 4:1 4:3 4:5 5:4 8:1 9:16 16:9 21:9 |
gemini-3-pro-image-preview |
1:1 2:3 3:2 3:4 4:3 4:5 5:4 9:16 16:9 21:9 |
| (default / other) | 1:1 3:4 4:3 9:16 16:9 |
1:4, 1:8, 4:1, 8:1 are flash-only. Do not use them with the pro model.
Image size
| Model | Supported values |
|---|---|
gemini-3.1-flash-image-preview |
512 0.5K 1K 2K 4K |
gemini-3-pro-image-preview |
1K 2K 4K |
| (default / other) | 1K 2K 4K |
512 and 0.5K are flash-only. Do not use them with the pro model.
Intent to Parameters
Translate clear user intent into CLI flags when the mapping is obvious.
Aspect-ratio guidance:
- Use
--aspect-ratio 9:16for requests such as phone wallpaper, vertical poster, story format, or other clearly vertical mobile outputs. - Use
--aspect-ratio 16:9for requests such as desktop wallpaper, presentation cover, horizontal banner, or other clearly widescreen outputs. - Use
--aspect-ratio 1:1for square social images, avatars, or other explicitly square outputs. - If the user already provides a specific ratio, pass it through directly.
- If the output shape is unclear, omit
--aspect-ratioand letyummycliuse its default.
Image-size guidance:
- Use
--image-size 4Kwhen the user explicitly asks for 4K or a clearly print-grade / high-resolution deliverable. - Use
--image-size 2Kwhen the user explicitly asks for 2K or a medium-resolution deliverable. - Use
--image-size 1Kwhen the user explicitly asks for 1K or a lightweight preview-sized result. - Use
--image-size 512or--image-size 0.5Konly when the user explicitly asks for a minimal-size output and the flash model is in use. - If the user does not explicitly request output size, omit
--image-sizeand letyummycliuse its default (1K).
Do not guess --image-size from general quality adjectives alone.
Output path guidance:
- If
--outputis omitted,yummycligenerates a default filename in the current working directory. Do not invent your own output filename unless the user explicitly provides one.
Examples
Prompt-only generation:
yummycli gemini nanobanana \
--prompt "A single ripe banana on a white plate, studio lighting, realistic photo"
Single-image edit:
yummycli gemini nanobanana \
--prompt "Turn this into a watercolor illustration" \
--input-image ./source.png
Multi-image reference edit:
yummycli gemini nanobanana \
--prompt "Blend these references into one polished poster illustration" \
--input-image ./subject.png \
--input-image ./background.jpg
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install yummy-gen-image - After installation, invoke the skill by name or use
/yummy-gen-image - Provide required inputs per the skill's parameter spec and get structured output
What is Yummy Gen Image?
Use when the user wants to generate or edit raster images with Gemini through yummycli, including prompt-only generation, single-image editing, and multi-ima... It is an AI Agent Skill for Claude Code / OpenClaw, with 143 downloads so far.
How do I install Yummy Gen Image?
Run "/install yummy-gen-image" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Yummy Gen Image free?
Yes, Yummy Gen Image is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Yummy Gen Image support?
Yummy Gen Image is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Yummy Gen Image?
It is built and maintained by yummysource (@yummysource); the current version is v1.1.0.