AI model APIs: 257 models for image, video, audio and text

The catalog has 257 models from 31 providers: 103 for video, 78 for image, 32 for audio and 44 for text. Picsart's own editing models, such as background removal, upscaling and instruction edits, are part of the same catalog and run on the same key as every other model.

Yes. One key reaches every model in the catalog, and the SDK and REST API send it as a bearer token. The CLI signs in through your browser instead, so there is no key to manage. Create your key in the console, then follow Authentication to set it up.

Change the model id. The TypeScript SDK uses the same generate() call for every model, and the REST API uses one endpoint with the same request and response shape. Only each model's parameters differ, and every model page lists them with copy-paste examples you can run right away.

Open any model to see its parameters, its rates for every setting and example requests. The playground on the same page runs the model once you sign in, and each run is billed at that model's published rate, so try one prompt on two or three models before you commit.

Yes, on models with an Advanced tab. Next to the unified API, that tab shows a compatibility API that takes the provider's original parameter names, unchanged. If your code already calls a provider such as Kling directly, it is the shortest way to move that code onto Picsart.

You pay per generation, with no seats and no subscription minimum. Video bills per second of output, images per generation, speech per 1K characters or per minute, and text per 1M tokens. The Pricing tab on each model's page lists every rate, and the SDK quotes the API cost of a call before you run it.

There are four: the type-safe TypeScript SDK, the REST API for any language, the gen-ai CLI for scripts and automation, and the MCP server, which lets agents such as Claude Code and Cursor run models directly. All four reach the same catalog from one account.