Generative AI | Run Image, Video, 3D and Audio Models | fal

fal is a generative media platform for developers focused on image, video, audio, 3D, and code models.
The platform offers 1,000+ production-ready generative media models accessible through a simple API.
Developers can use serverless GPUs and on-demand clusters to run, deploy, train, and fine-tune models.
fal’s serverless engine is designed for rapid inference with no GPU setup, cold starts, or autoscaler configuration.
The company provides dedicated compute for frontier research labs, including training and fine-tuning on latest NVIDIA hardware.
fal claims its inference engine is up to 10x faster for diffusion models and supports scaling to 100M+ daily inference calls.
Users can deploy private or fine-tuned models with one click and bring their own weights securely.
Pricing includes pay-per-output serverless usage or hourly GPU compute, with H100/H200/B200 starting at $1.89.
Enterprise features include SOC 2 compliance, single sign-on, private endpoints, usage analytics, and 24/7 priority support.
Customers cited include Canva, Perplexity, PlayAI, and Quora, with Poe reportedly using fal for 40% of its official image and video bots.

Your notes

Save this item to your library to add private notes.