Connect once via MCP. Your coding agent generates images, icons, video, voice and sound — and gets back a durable URL.
No setup · Any provider · Durable URLs
{
"mcpServers": {
"media": {
"command": "node",
"args": ["./apps/mcp/dist/index.js"],
"env": {
"MEDIA_API_KEY": "mk_••••••••"
}
}
}
}What you can generate
Everything an app ships with — what it looks like, and now what it sounds like.
Photorealistic, illustrated, or abstract — auto-optimized to WebP. Pick a model or let the engine choose.
App icons, favicons, splash screens, store screenshots and social cards. Every size plus a manifest.
Short clips, plus mascot animations rendered as transparent-background WebM.
Infinitely scalable SVG logos and icons, and Lottie JSON for looping UI micro-animations.
Narration, in-app announcements and multi-speaker dialogue. Ask for a voice by name — or design your own.
One call returns a whole interaction sound pack — tap, success, error, notification — plus music beds and loops.
Getting started
No SDK to install. No model juggling. Just an MCP config and a natural language prompt.
Add the MCP server config to your agent. One JSON block, one restart — and your agent has media superpowers.
Prompt naturally — "a hero image, dark theme, futuristic." The engine selects the right model, enhances your prompt, and generates.
A CDN-hosted asset URL lands in your agent's context. Stored in Cloudflare R2, served from the edge. Use it anywhere.
Under the hood
You write a prompt. The engine takes care of everything else.
fal.ai, OpenAI, Imagen, Replicate. Switch models without touching code — the registry handles routing.
Claude rewrites your prompt to match the model's strengths before generation. Better results, effortlessly.
Per-project style profiles. Set it once — every generated asset stays consistent across your project.
Generate N variants silently, score them, and return only the highest-rated one. Quality on autopilot.
Every asset stored in Cloudflare R2 and served from the global edge. Permanent links, zero expiry.
Email-OTP login, mintable API keys, and per-key usage tracking. Know exactly where every generation goes.
Ask once, get every clip your interaction layer needs — tap, success, error, notification — in one style, with a manifest that maps name to URL.
Generated clips arrive at wildly different volumes. WAV output is peak-normalized, and inaudible takes are regenerated automatically.
Say "narrator" instead of a voice id, or design a brand voice from a description and reuse it for every line your app speaks.
Pricing
Start free. Scale as you generate.
Free
Pro
Team
Connect Media Generator in under two minutes.
No credit card required · Free forever core