Skip to content

Which tool is right for me?

Picsart has one model catalog and several ways to reach it. The right surface depends on what you are doing, not on technical preference. This page helps you decide.


I want to explore models, try prompts, and see what's possible

Use: AI Playground (web or mobile)

The Playground is the fastest way to run a generation. No setup, no install, no config. Open a browser or the mobile app, pick a model, type a prompt, see the result. You can compare two models side by side, adjust parameters with sliders, and save results to your Drive.

If you are evaluating whether a model fits your creative brief, or iterating on a prompt to get it right, start here. Everything you learn in Playground transfers directly to the CLI, MCP, SDK, and REST API. The model ids and parameters are identical.

picsart.com/ai-playground


I work in Claude, ChatGPT, Cursor, or another AI assistant

Use: Skills or MCP

If you already spend your working day in an AI coding assistant or chat agent, you do not need to switch to Playground to generate media. Connect Picsart to your agent and generate directly from a conversation.

Skills are the faster path. Install once and ask in plain English: "Generate four product images on a white background using Flux 2 Pro." The agent handles everything: model selection, parameter construction, command execution, result delivery.

MCP gives you tool-level access. Use it when you want the agent to reason about pricing before generating, validate a payload, chain background removal after generation, or manage Drive files. All in one agent turn.

Typical workflow: validate the prompt in Playground, connect the agent, then generate at scale from the conversation.


I need to run generations on a schedule or as part of a script

Use: CLI

The gen-ai CLI is built for automation. Every generation flag is scriptable, output is JSON-ready, and it runs in any environment where you can run a shell command: cron, Docker, GitHub Actions, GitLab CI, a plain shell script.

Examples:

bash
# Scheduled catalog refresh via cron
0 6 * * 1 gen-ai batch run /home/user/weekly-catalog.yaml

# Generate assets as a CI step
gen-ai generate -m flux-2-pro -p "product hero" --script | jq '.results[0].url'

# Process a folder of images
gen-ai generate -m wan-2.7-i2v -p "subtle animation" --input-dir ./stills/

You can also quote costs before a batch run, resume failed jobs, and write results directly to Picsart Drive.


I know my way around the terminal but I am not writing app code

Use: CLI

If you work in the terminal, write shell scripts, or run ad hoc jobs but are not building a product or service, the CLI is your surface. You get the full model catalog as a single command, JSON output you can pipe anywhere, and no boilerplate.

bash
gen-ai generate -m flux-2-pro -p "hero shot, soft light" --ar 16:9
gen-ai models --mode image | grep recraft

The CLI authenticates with OAuth: one gen-ai login in the browser. No API key setup required.


I am building an application or website that needs generative AI

Use: SDK (Node.js/TypeScript) or REST API (any language)

If you are writing code to power a product — a web app, a mobile app, a backend service, a SaaS feature — choose based on your stack.

Node.js or TypeScript

Use @picsart/ai-sdk. It is type-safe, model-aware, and handles async polling automatically. One generate() call covers all 174 models. Results can auto-save to Picsart Drive.

bash
npm install @picsart/ai-sdk

Auth is an API key (not the CLI's OAuth flow). Retrieve it from picsart.com/settings.

SDK guide

Python, Ruby, Go, or any other language

Use the REST API directly. HTTP POST to https://api.picsart.com/gw-v2. Same model ids and parameters as every other surface. Auth is an API key from picsart.com/settings.

REST API guide · API Reference


I am not sure — I just want to try it

Use: AI Playground

Start at picsart.com/ai-playground. No account setup other than a Picsart login. Pick any model from the catalog, write a prompt, and generate. You can decide later whether you want to automate it via the CLI or connect it to your agent.


Side-by-side comparison

SituationSurface
Trying a new model for the first timeAI Playground
Comparing two models on the same promptAI Playground
Iterating on a prompt until it's rightAI Playground
Generating inside Claude or ChatGPTSkills or MCP
Generating inside Cursor or WindsurfSkills or MCP
Batch-generating a product catalogCLI
Scheduling weekly or daily asset creationCLI
Building a CI/CD pipeline with media generationCLI
Ad hoc terminal work, scripting, one-off jobsCLI
Building a Node.js/TypeScript appSDK
Building in Python, Ruby, Go, or any other languageREST API
Running generations from application codeSDK or REST API
Not sure — just exploringAI Playground

Can I use more than one?

Yes. The surfaces are complementary. A common pattern:

  1. Experiment in Playground to find the right model and prompt.
  2. Connect MCP or Skills so your AI agent can generate the same thing without switching windows.
  3. Set up a CLI batch job to run nightly for catalog updates.
  4. Call the REST API from your product for on-demand user-triggered generation in any language.
  5. Use @picsart/ai-sdk in your Node.js service for type-safe, auto-polling model calls.

All five use the same account, the same model ids, and the same credit balance. The CLI and agents use OAuth; the SDK and REST API use an API key.

Built on @picsart/ai-sdk · gen-ai CLI · Picsart MCP · Skills