Introduction
Picsart AI Playground
Picsart AI Playground brings together AI models under one account: one credit balance, no separate subscriptions, no per-provider API keys. You can generate images, video, and audio using models like Flux, Seedance, Veo, Kling, ElevenLabs, Recraft, Seedance, and others, and switch between them freely.
The Playground is available on web and mobile. For most users and most tasks, this is the right place to work. You pick a model, write a prompt, see results, iterate. If you are generating at small scale, exploring what is possible, comparing model outputs, or doing prompt engineering, the Playground web or mobile app is the most efficient tool. No setup, no code, no configuration.
These docs cover the programmatic and agent-native interfaces to the same platform: the gen-ai CLI, Skills, the MCP server, the TypeScript SDK, and the REST API. Each one is an extension of the Playground, not a replacement for it.
The access layers
Picsart has six ways to reach its model catalog beyond the web and mobile app. Each one fits a different working context.
AI Playground (web and mobile)
Start here. Experiment with models, compare outputs side by side, tune prompts, and see what each model does well. Good for: solo creative work, model evaluation, prompt engineering, and anything you would otherwise do in a browser or on your phone. If the scale is small and speed-of-delivery matters more than automation, Playground is the most direct path.
Skills and MCP: for your AI agent
Once you have validated a prompt or workflow in Playground, connect Picsart to the AI agent or coding assistant you already work in: Claude Code, ChatGPT, Cursor, Windsurf, Codex, or VS Code Copilot.
When the connection is active, you can generate images, animate a video, remove a background, or synthesize audio without leaving your agent conversation. The agent handles the model selection, parameter construction, and result retrieval. You stay in Claude or ChatGPT and issue natural-language instructions: "Generate three hero images for this brief in 16:9."
Skills are pre-built instruction bundles that give the agent knowledge about Picsart's models and workflows. Install once and describe tasks in plain English. MCP is the protocol that exposes every catalog tool directly, useful when you want the agent to reason about cost, validate parameters, or chain multiple operations. The Picsart MCP server is hosted, so you only add its address to your agent and sign in; there is nothing to install.
Skills guide · MCP Quickstart · Integrations
CLI: terminal, scripts, and scheduled work
The gen-ai CLI exposes the full catalog as a single terminal command. It is pipe-friendly, scriptable, and designed for automation. Useful when:
- You want to run generations on a cron schedule (daily catalog refresh, weekly campaign assets).
- You are building a shell pipeline that generates, then processes, then uploads.
- You need CI/CD integration: generate assets as part of a build step.
- You want to batch-process a folder of product images or run a manifest of mixed generation jobs.
gen-ai generate -m flux-2-pro -p "product on white background" -n 4
gen-ai batch run catalog.jsonCLI Quickstart · Batch and Automation
SDK: Node.js and TypeScript applications
If you are building a Node.js or TypeScript application, @picsart/ai-sdk is the fastest path. It is type-safe, model-aware, and handles async polling automatically. The SDK 6.29.0 catalog covers 247 models. Results can auto-save to Picsart Drive. Auth is an API key (not OAuth), retrieved from your account settings.
npm install @picsart/ai-sdkREST API: any language
If you are building in Python, Ruby, Go, PHP, or any other language, call the REST API directly. HTTP POST to https://api.picsart.com. Same model ids and parameters as every other surface. Auth is an API key from your account settings.
REST API guide · API Reference
How the surfaces relate
They share an account and credit balance. The SDK 6.29.0 catalog covers 247 models from 32 providers. Model availability and parameters depend on the installed SDK, CLI, or server version. Each surface signs in its own way: the CLI and Skills use gen-ai login (Skills run the CLI internally), the MCP server is hosted by Picsart and your agent signs you in through your browser with nothing to install, and the SDK and REST API use an API key from your account settings.
| Surface | Who it is for | Typical use |
|---|---|---|
| AI Playground (web / mobile) | Any user | Exploration, iteration, small-scale creative work |
| Skills | Agent users (Claude, ChatGPT, Cursor, Windsurf) | Generate inside your agent in plain English |
| MCP | Agent users who want tool-level control | Chain tools, inspect cost, validate before generating |
| CLI | Developers and power users | Automation, cron, batch, CI/CD pipelines |
| SDK | Node.js/TypeScript developers | Application integration, type-safe model calls, Drive auto-save |
| REST API | Developers in any language | Application integration from Python, Ruby, Go, or any HTTP client |
Most people move between these surfaces naturally. You experiment in Playground, validate a prompt, then wire it into an agent or a script. The platform is the same at every level.
What you can generate
- Image — text-to-image, image editing, inpainting, style transfer, background removal/replacement, upscaling, and vector/SVG output. 73 image models.
- Video — text-to-video, image-to-video, video-to-video editing, and clip extension. 99 video models.
- Audio — text-to-speech, music, sound effects, voice design, dubbing, and speech-to-speech. 32 audio models.
- Text analysis — describe, caption, OCR, and summarize images and video using Claude, GPT, or Gemini. 42 text models.
Browse the full catalog in the Model Reference, filter live in the Playground, or query from the terminal:
gen-ai models --mode video
gen-ai models --provider google
gen-ai models info seedance-2.0Next steps
Not sure which surface to start with? Read Which tool is right for me?.
- Installation: install the CLI, add Skills, or connect the MCP server.
- Authentication: sign in once.
- CLI Quickstart · MCP Quickstart · Skills
- SDK · REST API
- Integrations: per-agent setup guides for Claude Code, Cursor, Windsurf, ChatGPT, Codex, VS Code.
Structured evaluation is also available through the SDK: 1 JSON model uses ai.run() and returns named answers. See JSON & evaluation.