Cookbook
Real recipes for building with AI models — single-model tricks, multi-model pipelines, and the API code to ship them yourself.
The EvoLink Cookbook is for AI creators who work in Codex and Claude Code. Each recipe is a real build — chain image → video → upscale, or audio + video — with the exact prompts and copy-able API calls behind it. No one-click templates: you get the building blocks and wire them together through one unified API.
Single-model recipes
What each model is genuinely good at — with prompts and real examples.
Multi-model combos
Chain models into one result — image → video → upscale, audio + video.
Prompts & examples
Copy-paste starting points you can adapt to your own project.
How to build it
Step-by-step implementation with the EvoLink API in Codex or Claude Code.
Blender → Video
Model a scene in Blender, paint a first frame, and turn it into a finished 4K clip with Seedance — all inside your agent.
View recipeSeed Audio · Dialogue
Generate multi-character dialogue, sound effects, and music in a single pass.
Coming soonMore recipes coming soon—Seedance · Street-Style Video · Veo 3.1 Fast · Commercial Shots · Nano Banana 2 · Image Editing
Pick a recipe
Choose a recipe and follow it end to end — the models, prompts, and steps are all laid out.
Install & get your key
Set up the tools, then grab your EvoLink API key — one key powers every model in the pipeline.
Prompt your agent
Describe what you want. Your agent runs the whole pipeline for you — no playground needed.
One API key
Every model in the pipeline runs on a single EvoLink key.
Runs in your agent
Cursor, Claude Code, Codex — any MCP client. No playground.
Cheapest provider, auto
Smart routing picks the lowest-cost provider for every call.
Production ready
99.9% uptime with automatic failover across providers.
What is a recipe here?
An end-to-end build — the models, prompts, and steps to take an idea to a finished asset, run inside your AI agent. Not a one-click template; you implement it yourself.
Do I need to use a playground?
No. Workflows run directly inside your agent (Cursor, Claude Code, Codex) over MCP.
Do I need a separate key for each model?
No. One EvoLink API key powers every model in the workflow — image, video, audio, and upscaling.
How much does it cost?
Pay-as-you-go by usage. Smart routing automatically picks the cheapest provider for each call.
Can I change the models or steps?
Yes. Recipes are a starting point — swap models, tweak prompts, and adapt the steps to your project.
Is this a downloadable template or one-click workflow?
No. Each recipe is an approach you implement yourself with the EvoLink API in your agent — not a node graph or template to import.
Do I need to know how to code?
If you can run commands in an agent like Codex or Claude Code, you can follow a recipe. Many steps are copy-paste.
Which models can I use?
Every model on EvoLink — image, video, audio, and LLMs — including Seedance, Midjourney, Veo, Nano Banana, Seed Audio, Topaz and more.
Can I combine multiple models in one recipe?
Yes — that's the whole point. Chain models like image → video → upscale, or audio + video, all through one API.
How is this different from the blog or prompt pages?
The blog covers news and launches; prompt pages are prompt galleries. Recipes are end-to-end builds — models, prompts, and steps to a finished result.
Is EvoLink compatible with the OpenAI SDK?
Yes. EvoLink is 100% OpenAI-SDK compatible — point your existing code at the EvoLink base URL and your key.