Replicate

What it does

Replicate is a cloud API for running machine learning models. You can run community models with one line of code, fine-tune image models with your own data, or deploy custom models packaged with Cog. Compute scales automatically and billing is based on how long your code runs. Sign in with GitHub to get started, and use the documentation for Node.js, Python and HTTP examples.

The site lists image, video, speech, music and language models, including official models from several providers. Fine-tuning is demonstrated for image models such as FLUX. Custom deployment uses Cog, an open-source tool that generates an API server. Pricing is listed per second for CPU and several Nvidia GPUs. The docs cover models, predictions, deployments, webhooks, organizations, billing and API tokens.