Images · Video

fal

Developer platform for generative media inference and serverless GPUs, hosting 1,000+ image, video, audio, and 3D model APIs including FLUX, MiniMax H3, Kling, Seedance, and ElevenLabs with prepaid output-based billing and custom GPU deployment.

Pay-as-you-go

Pay-as-you-go

Pay as you go

2 concurrent requests initially (auto-scales up to 40 based on 4-week credit purchases), $0 monthly base fee, billed per output or Serverless GPU second from prepaid credits

Sign up at fal.ai, generate a FAL_KEY in the dashboard, top up prepaid credits, and call hosted models via Python/JS SDKs, REST queue endpoints, genmedia CLI, MCP server, or the web fal Agent.

Chinese supportedLicensed commercial useAPI available

Overview & quick start

Official docs

1,000+ hosted generative media APIs

Billed per image, megapixel, video second, or 1,000 characters

Official MCP server integration (Claude Code / Cursor / Codex)

Supports OAuth or API key setup in coding and chat agents

Serverless GPU custom deployment (H100/H200/B200/B300)

Requires account deploy approval; H100 from $2.49/hr billed per second

Best for

Developers and engineering teams integrating low-latency image, video, speech, and 3D generation APIs, or deploying custom Python and Docker models on per-second H100/H200/B200 serverless GPUs.

How to get started

Sign up at fal.ai, generate a FAL_KEY in the dashboard, top up prepaid credits, and call hosted models via Python/JS SDKs, REST queue endpoints, genmedia CLI, MCP server, or the web fal Agent.

Platforms & clients
Model & Platform APIs · fal CLI & genmedia CLI · Web Dashboard, Playground & fal Agent · API · MCP
Region & network
GlobalAvailable globally excluding sanctioned jurisdictions; billed in USD via self-serve credit top-ups or enterprise invoicing.
Payment & sign-in
Credit / debit card / InvoiceSign in: GitHub / Google / SSO
Data & privacy
Enterprise: Not used for training;Pay-as-you-go: Not used for trainingGenerated CDN media stored for at least 7 days by default (configurable per request via X-Fal-Object-Lifecycle-Preference); request payload storage can be disabled per request via X-Fal-Store-IO: 0; fal Agent keeps chats Forever by default with 1h-1y options on Enterprise. · Compliance: SOC 2
Rate limits
Default 2 concurrent IN_PROGRESS requests for new accounts; auto-scales up to 40 with 4-week credit purchases; unlimited queue depth; custom >40 on Enterprise.