# Image Generation

> Generates images using Stable Diffusion with checkpoint switching, LoRA support, and self-portrait mode. Use when the user requests pictures, artwork, selfies, or portraits.

- Skill: `rotoslider/image-generation` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add rotoslider/image-generation`
- Raw SKILL.md: https://api.skillmd.com/api/skills/rotoslider/image-generation/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: rotoslider (https://skillmd.com/u/rotoslider)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/rotoslider/image-generation

---


# Image Generation

## When to Use
- User requests image of themselves → `generate_image` with `self_portrait: true`
- User requests general image → `generate_image` with `self_portrait: false`
- Save generated image to project → `save_generated_image` with the `imageId` from generate_image result

## Self-Portrait Detection
Auto-detected from phrases: "selfie", "picture of me/you/yourself", "what do you look like"

## Size/Aspect Parameters
- size: "small" (768px), "medium" (1024px), "large" (1536px), "x-large" (1856px)
- aspect: "portrait" (3:4), "portrait-tall" (9:16), "square" (1:1), "landscape" (16:9), "wide" (21:9)

## Important
- Checkpoint switching is serialized via image generation lock to prevent race conditions
- LoRA models injected as `<lora:name:weight>` in prompt
- Dimensions must be divisible by 32
- Auto-detects checkpoint type from name (pony, flux, other) for CFG settings