MOSS-Image Workflow

MOSS-Image Workflow

Generates images from text with the MOSS-Image Base and SFT models on a node-canvas Gradio page, with sampling controls and an examples strip that restores its settings. For people trying MOSS-Image in the browser.

Made by OpenMOSS-Team on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: MOSS-Image Workflow is a Gradio page that runs MOSS-Image text-to-image generation as a node canvas, built on the gr.Workflow feature. It is a workflow version of the official MOSS-Image demo. Both the Base and SFT checkpoints load at startup. You can open it in the browser with nothing to install.

Who it suits: It suits people who want to try MOSS-Image with every sampling control exposed, and developers curious about Gradio's workflow canvas.

Industry | Who uses it | How they use it | Example
Design | Designers | Generate pictures from a written prompt | Type a prompt and pick the Base model
Illustration | Artists | Compare the Base and SFT models | Run the same prompt with each model
Research | AI researchers | Tune sampling settings step by step | Change Steps and CFG and compare results
Software | Python developers | Study a node-canvas app built on gr.Workflow | Read gradio_app.py and workflow.json

Skip it if you need to add your own showcase images to a gallery, since that feature was not ported.

Where to use it:

• A web browser, on the Space's page at Hugging Face, which runs on ZeroGPU
• Your own Windows, macOS or Linux computer with Python 3.12

Use case | Who | What they get back
Generate pictures from text | Designers | One or more images in a gallery
Reuse a showcase recipe | Artists | A form filled in from one of the example images
Explore sampling controls | Researchers | Results that change with each setting

How to set it up: The online page needs nothing. A local copy needs Python 3.12 and the packages in requirements.txt.

• Python 3.12 and pip
• Python packages from requirements.txt
• Windows, macOS or Linux

1. Install the Hugging Face hub library.
2. Download the Space's files into a folder named MOSS-Image.
3. Move into that folder.
4. Install Gradio 6.30.0, then the packages from requirements.txt.
5. Start app.py.

pip install -U huggingface_hub
hf download OpenMOSS-Team/MOSS-Image --repo-type space --local-dir MOSS-Image
cd MOSS-Image
pip install gradio==6.30.0
pip install -r requirements.txt
python app.py

How to use it: Open the Space's page, or run the last command above for a local copy and open the address it prints. Each sampling control is a reference node wired into the generate step.

python app.py

Tool | What it does
Model | Chooses the base or sft checkpoint
Prompt and Negative Prompt | The text to draw and the text to steer away from
Width and Height | Image size, in multiples of 16
Steps, CFG, Shift and Seed | Sampling settings, where a seed of -1 means random
Images | How many pictures to make, shown in a gallery
Max Text Length | A limit on prompt length, from 1 to 4096
Skip Layers and SLG settings | Layer skipping options, where 22 means negative-branch skipping and empty means off
Examples strip | Showcase images that fill the form with their saved settings when clicked

Example: You click one of the showcase images in the examples strip. The form fills with that image's prompt, model, size and sampling settings, and you run it to get your own version.

For newcomers and experts: Click an example in the strip, run it, then change one setting at a time. That shows what each control does.

The app uses gr.Workflow in Gradio 6.30.0, and its look is applied through launch options for the theme, CSS and head. The canvas view is left untouched. This version runs the eager backend and leaves out ZeroGPU AOT compilation, so it may be slower than the original. Showcase add and delete with live refresh was not ported, because the generated App view has no upload-to-gallery mechanism.

Open the page
Set or restore the controls
The generate node runs the chosen model on a ZeroGPU
Images appear in the gallery

Good to know:

• Not ported: Adding or deleting showcase images, and ZeroGPU AOT compilation.
• Hardware: The Hugging Face copy runs on ZeroGPU hardware.
• License: apache-2, with a LICENSE-CODE file and notices in the repository.
• Cost to run: none named.
• Data it reaches: Your prompts and settings. The examples strip loads showcase images and their saved settings from outside the page.

On Hugging Face, 13 November 2026

• Likes: 193
• Built with: gradio
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/OpenMOSS-Team/MOSS-Image
• Files: https://huggingface.co/spaces/OpenMOSS-Team/MOSS-Image/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.