Generate AI images & videos on your GPU, in your agent, or via API.
One catalog — 45+ image models and 3 video models — three ways in: a free desktop studio that runs models locally, an MCP server for Claude and any agent, and a REST API with credits or crypto pay-per-use.
Imference Desktop
Free · Win & macOSA full generation studio. Run seven model families locally on your GPU — private, unlimited, $0 per image — or switch to cloud for the full catalog, video included. No account, no subscription — download and generate.
Download free →
Imference MCP
New · npmGive Claude Code, Claude Desktop or Cursor the power to make images. Your agent picks the model, prompts it the way its card recommends, sees the result and iterates.
Set up in one line →
Imference API
REST · credits or x402The engine under both. One endpoint for every model, prepaid credits or anonymous USDC pay-per-use. From $0.003 per image.
curl -X POST https://imference.com/generate \
-H "Authorization: Bearer $KEY" \
-d '{"model":"z-image-turbo","prompt":"…"}'
Read the docs →
Imference Desktop — a real studio on your own hardware
Run models locally on your own GPU — private and free — or generate in the cloud with credits or x402. One interface, one gallery, no subscription.
- Seven local model families — SDXL, FLUX, Z-Image & more — on NVIDIA, AMD or Apple Silicon
- Cloud mode uses the same API — credits or x402 pay-per-use
- Image-to-image + metadata-rich gallery built in
Local
Your GPU
Private · $0 per image
Cloud
imference.com
No GPU required
Prompt
cinematic portrait, soft rim light…
Generate images without leaving your agent
imference-mcp gives any MCP client eleven tools over the full Imference catalog. Ask for “a cover image — pick the best model”: the agent compares model cards, applies the recommended prompt style, sees the finished image in the conversation, and iterates.
- The agent sees every image it generates and self-corrects
- Model cards with exact parameter bounds — the same list the API validates
- Pay with an API key, or per-image from a wallet (x402) with spending caps
One command:
claude mcp add imference -e IMFERENCE_API_KEY=your-api-key -- npx -y imference-mcp
→ list_models · 45 image models
→ get_model("hassaku-anima")
→ generate(format_code: portrait-large)
The agent sees the image it made — and fixes its prompt if it missed.
Everything runs on one API
The desktop app's cloud mode, the MCP server and your own code all call the same endpoint — same catalog, same prices, two payment rails.
Unified endpoint
Swap models with a single field. Keep your integration stable as you iterate.
Credit-based billing
Purchase credits via Stripe. Ideal for teams, dashboards, and predictable monthly spend.
x402 pay-per-use
Anonymous on-chain payments (USDC on Base). Perfect for AI agents.
Choose your payment method
Start with credits for tracking, or go wallet-first with x402 for pay-as-you-go.
Credit-based API
Best for predictable billing
Purchase credits via Stripe and use your API key for authenticated requests. Great for dashboards, usage tracking, and teams.
- Per-model pricing in credits
- API key authentication
- Ideal for production workloads
Pay-per-use (x402)
No account required
Pay with USDC via your Web3 wallet. Anonymous, on-chain payments powered by x402.
- USDC on Base network
- Anonymous pay-as-you-go
- Perfect for embeds & demos
Available models
The shared catalog behind the app, the agent and the API — realism, illustration, anime, plus hosted providers for image/video.
Loading models…
Pricing that stays simple
Use credits for predictability, or pay per image with your wallet.
Credit-based
$5
for 5,000 credits
1 credit = $0.001
cost per generation varies by model
Pay-per-use (x402)
from $0.003
per generation
price varies by model — same rate as credits
USDC on Base network
Pick your door
Free and local with the desktop app, one command away in your agent, or straight HTTP with the API.