236 AI models. One API key.

One API for generative AI across four modalities — image, video, audio and text — plus Picsart’s own image editing models. You get a single API key, a type-safe TypeScript SDK, a REST endpoint, a CLI and an MCP server, all reaching the same catalog of 236 production-ready models from 31 providers.

Frontier models from 31 providers, including Google Veo and Lyria, OpenAI, Anthropic Claude, Flux, ByteDance Seedance, MiniMax Hailuo, Kling, Luma Ray, Runway, Pika, Recraft, Ideogram, ElevenLabs, HeyGen, Topaz and Alibaba Wan — alongside Picsart’s own generation and editing models. Every model is browsable with its parameters and rates on the models pages.

Yes. Background removal is a first-party Picsart model (picsart-sod-v8-2) that returns a precise cutout with the alpha channel preserved. There is also picsart-change-bg to replace a background from a text prompt, picsart-enhance for upscaling with face enhancement, and picsart-qwen-image-edit for instruction-based edits. These are the same models behind Picsart’s own editing tools.

Yes — video is the largest part of the catalog. You can generate video from text or animate a still image, and the same generate() call works across every video model, so switching from one to another is a change of model id rather than a rewrite.

You pay per generation — no seats, no subscription minimum and no tiered discount puzzles. Every model publishes its own rate, and the SDK can quote the cost of a call before you run it, so spend is predictable before a job starts rather than after it.

Two differences. First, breadth of kind rather than count: most gateways cover one domain — OpenRouter is text, fal and Replicate are media, Higgsfield is video — while this API covers image, video, audio and text on one key. Second, and more useful in practice, it does not stop at generation: background removal, cutouts, enhancement and instruction edits are first-party models on the same endpoint, so a pipeline that generates and then edits never leaves the API.

Yes. An MCP server ships with the platform, so Claude Code, Cursor and any MCP-compatible client can reach the full catalog with your key. There is also a gen-ai CLI for scripting and automation from a terminal.

The @picsart/ai-sdk package gives TypeScript and JavaScript full types for every model and parameter, including editor autocomplete on model ids. Any other language calls the REST API over plain HTTP with the same parameters and the same responses.