Home/ IMAGE/ Jaaz: An Open-Source, Local-First Canvas for AI Images and Video
IMAGE

Jaaz: An Open-Source, Local-First Canvas for AI Images and Video

Jaaz is an open-source Canva AI alternative and local-first creative canvas for AI images and video, with an AI agent that can use local ComfyUI or cloud models. License unverified.

Published Sep 15, 2026 · 3 min read
Jaaz local-first open-source AI image and video canvas
TL;DR: Jaaz is an open-source, local-first creative canvas for AI images and video โ€” a privacy-minded alternative to Canva whose AI agent can call local ComfyUI models or paid cloud APIs. License terms are unverified as of publication.

Jaaz is a desktop creative app built around an infinite canvas and a chat agent. You sketch, arrange and describe; the agent calls image and video models to fill in the rest. Its maintainers position it as a privacy-minded, open-source alternative to Canva and Manus that can run against local models as well as paid APIs.

Repository11cafe/jaaz
Maintainer11cafe
Primary languageTypeScript (Python backend)
GitHub stars6,638 (as of 15 September 2026)
Last push2 March 2026
License statusUnverified โ€” see below

What it does

The README describes four main pieces:

  • One-prompt image and video generation, routed to models the README names as GPT-4o, Midjourney, Veo 3, Kling and Seedance, with automatic prompt optimisation and multi-turn refinement.
  • “Magic Canvas” and “Magic Video”: sketch, point with arrows or annotate a frame, and the agent interprets the layout instead of a written prompt.
  • An infinite canvas for storyboarding, with linked layouts and a media library.
  • An agent that inserts objects, transfers styles and, per the README, works with both local ComfyUI and cloud models, including a fully offline or hybrid setup using Ollama.

Who it’s for

  • Designers and content creators who prefer arranging things visually to writing long prompts.
  • People who already run ComfyUI or Ollama locally and want a friendlier front end on top.

Getting started

Desktop builds for Windows and macOS are distributed from the project’s GitHub releases and jaaz.app. The README says cloud API models are accessed by logging in on the app’s homepage with a paid plan. For Linux or local builds, it documents a manual install that needs Python 3.12 or newer: build the React front end with npm install --force and npx vite build, then install the server requirements and run python main.py. The README does not state hardware requirements; those depend on which local models you connect.

License status

Unverified. GitHub could not match this project’s license file to a standard open-source license (it reports NOASSERTION). The Signal has not reviewed the terms. The README also advertises a separate enterprise edition with “commercial licensing”, which suggests the open repository and the paid product may be licensed differently. Read the license file before any commercial use, redistribution or modification.

Limitations and open questions

  • Many of the headline models (GPT-4o, Midjourney, Veo 3, Kling) are paid cloud services, so “local-first” applies only if you connect local models yourself.
  • The manual-install instructions clone a repository named 11cafe/localart rather than 11cafe/jaaz; worth checking before you follow them.
  • The last push was in March 2026, and the enterprise promotion in the README carries a 2025 deadline, so parts of the README may be out of date.
  • We have not installed or tested Jaaz; features described here come from its README.

Sources

Related: more repo profiles ยท ComfyUI-LTXVideo, for the ComfyUI side of video generation.

This profile was drafted with AI assistance from the project’s public repository and reviewed by The Signal’s editors before publication. See our AI Disclosure.

The Signal newsletter

Keep getting this

One edition a week on open models, local setups and the tools around them.

Read the latest issue

Email delivery opens once the newsletter platform is connected.

Scroll to Top