What it is
qwen image studio is a tool by GitHub user janishar for running the open image model Qwen-Image-2.1 locally on an Apple Silicon Mac, for text-to-image and multi-image editing. It has two parts: a command-line tool and a local web studio in a browser tab that queues renders and streams their progress. Qwen in the name is the model it runs, released by the Qwen team; this repo is the author's own third-party tool. What sets it apart is how it saves memory: when asked, a 9B enhancer rewrites your prompt and is freed before the image model loads, so the two never sit in memory together.
Highlights:
- The README says rendering happens on your machine and the studio sends nothing off it
- reference-image editing and transparent backgrounds are supported
- an optional faster Turbo model renders in 8 steps
- sizes follow the aspect ratio, with 1:1 at 2048×2048.
How to use:
- First run uv sync and download the model weights.
- Then render from the command line, for example uv run qwen-image-2-1 "A neon shop sign that reads QWEN" -m ~/models/Qwen-Image-2.1.
- Finally, for the web studio, install the author's helmstudio launcher and install and launch it from there.
Who it's for:
Creators and developers with a high-memory Apple Silicon Mac who want to run an open image model locally.
Before you start:
- The author built and tested it on an Apple Silicon Mac with 64 GB of unified memory; CUDA and CPU options exist but are untested.
- You need uv, git and Hugging Face's hf CLI; the image model weights are about 30 GB from Hugging Face, and each prompt enhancer about 18 GB.
- The web studio runs only through helmstudio, has no authentication and binds to your own machine by default; exposing it beyond that is on you.
- If you use these features, weigh the safety risks yourself.
- AI Inspiration World is not responsible for any consequences.
- The tool is free; Turbo is under the Qwen Research License, and the README doesn't state the main model weights' license.
- The repo itself is MIT.
Based on the project's public documentation. Not tested by AI Inspiration World.