🎨 Image Generation — Image工具
v1.0.0Create AI images with GPT Image, Gemini Nano Banana 图像工具, FLUX, Imagen, and top providers using prompt engineering, style control, and smart editing.
详细分析 ▾
运行时依赖
版本
- Initial release of the AI Image Generation skill, providing modern AI image creation and editing across multiple providers. - Includes updated 2026 benchmark-guided model selection, with clear guidance on best practices and common pitfalls. - Supports prompt engineering, style control, and automated model alias resolution for consistent results. - Documents privacy practices and all external endpoints used. - Adds migration guidance, troubleshooting, and links to related skills for advanced workflows.
安装命令 点击复制
技能文档
Setup
On first use, read setup.md.
当...时 到 使用
User needs AI-generated visuals, edits, or consistent image sets. Use this skill to pick the right model, write stronger prompts, and avoid outdated model choices.
Architecture
User preferences persist in ~/image-generation/. See memory-template.md for setup.
~/image-generation/
├── memory.md # Preferred providers, project context, winning recipes
└── history.md # Optional generation log
Quick Reference
| Topic | File |
|---|---|
| Initial setup | setup.md |
| Memory template | memory-template.md |
| Migration guide | migration.md |
| Benchmark snapshots | benchmarks-2026.md |
| Prompt techniques | prompting.md |
| API handling | api-patterns.md |
| GPT Image (OpenAI) | gpt-image.md |
| Gemini and Imagen (Google) | gemini.md |
| FLUX (Black Forest Labs) | flux.md |
| Midjourney | midjourney.md |
| Leonardo | leonardo.md |
| Ideogram | ideogram.md |
| Replicate | replicate.md |
| Stable Diffusion | stable-diffusion.md |
Core Rules
1. Resolve aliases 到 official 模型 IDs 第一个
Community names shift quickly. Before calling an API, map the nickname to the provider model ID.
| Community label | Official model ID to try first | Notes |
|---|---|---|
| Nano Banana | gemini-2.5-flash-image-preview | Common nickname, not an official Google model ID |
| Nano Banana 2 / Pro | Verify provider docs | Usually a provider preset over Gemini image models |
| GPT Image 1.5 | gpt-image-1.5 | Current OpenAI high-tier image model |
| GPT Image mini / iMini | gpt-image-1-mini | Budget/faster OpenAI variant |
| FLUX 2 Pro / Max | flux-pro / flux-ultra | Many platforms rename these SKUs |
2. Pick models 由 task, 不 由 hype
| Task | First choice | Backup |
|---|---|---|
| Exact text in image | gpt-image-1.5 | Ideogram |
| Multi-turn edits | gemini-2.5-flash-image-preview | flux-kontext-pro |
| Photoreal hero shots | imagen-4.0-ultra-generate-001 | flux-ultra |
| Fast low-cost drafts | gpt-image-1-mini | imagen-4.0-fast-generate-001 |
| Character/product consistency | flux-kontext-max | gpt-image-1.5 with references |
| Local no-API workflows | flux-schnell | SDXL |
3. 使用 benchmark tables 作为 dated snapshots
Benchmarks drift weekly. Use benchmarks-2026.md as a starting point, then recheck current rankings when quality is critical.
4. Draft cheap, finish expensive
Start with 1-4 low-cost drafts, pick one, then upscale or rerender only the winner.
5. Keep fallback chain
If the preferred model is unavailable, fallback by tier: 1) same provider lower tier, 2) cross-provider equivalent, 3) local/open model.
6. Treat DALL-E 作为 legacy
OpenAI lists DALL-E 2/3 as legacy. Do not use them as default for new projects.
Common Traps
- 使用 vendor nicknames 作为 模型 IDs -> API errors 和 wasted retries
- Assuming "Nano Banana Pro" 或 "FLUX 2" universal IDs -> provider mismatch
- Copying 旧的 DALL-E prompt habits -> weaker 输出 vs modern GPT/Gemini image models
- Comparing text-到-image 和 image-editing scores 作为 如果 它们 是 相同 benchmark
- Optimizing every draft 在 max quality -> cost spikes 没有 quality gain
Security & Privacy
Data leaves machine:
- Prompt text
- Reference images 当...时 editing 或 样式 matching
Data stays local:
- Provider preferences 在...中
~/image-generation/memory.md - 可选 local history file
skill 做 不:
- Store API keys
- 上传 files outside chosen provider requests
- Persist generated images unless 用户 asks 到 保存 them
External Endpoints
| Provider | Endpoint | Data Sent | Purpose |
|---|---|---|---|
| OpenAI | api.openai.com | Prompt text, optional input images | GPT Image generation/editing |
| Google Gemini API | generativelanguage.googleapis.com | Prompt text, optional input images | Gemini image generation/editing |
| Google Vertex AI | aiplatform.googleapis.com | Prompt text, optional input images | Imagen 4 generation |
| Black Forest Labs | api.bfl.ai | Prompt text, optional input images | FLUX generation/editing |
| Replicate | api.replicate.com | Prompt text, optional input images | Hosted third-party image models |
| Midjourney | discord.com | Prompt text | Midjourney generation via Discord workflows |
| Leonardo | cloud.leonardo.ai | Prompt text, optional input images | Leonardo generation/editing |
| Ideogram | api.ideogram.ai | Prompt text | Typography-focused image generation |
Migration
If upgrading from a previous version, read migration.md before updating local memory structure.
Trust
This skill may send prompts and reference images to third-party AI providers. Only install if you trust those providers with your content.
Related Skills
Install withclawhub install if user confirms:
image-编辑- Specialized inpainting, outpainting, 和 mask workflowsvideo-generation- Convert image concepts 进入 video pipelinescolors- Build palettes 对于 visual consistency 穿过 assetsffmpeg- Post-process image sequences 和 exports
Feedback
- 如果 useful:
clawhub star image-generation - Stay updated:
clawhub 同步
免费技能或插件可能存在安全风险,如需更匹配、更安全的方案,建议联系付费定制