Wet-Ink

Wet-Ink

Animates a decoding method for diffusion language models in your browser, where you change a window size and watch tokens stay editable or freeze. For researchers and students of language model decoding.

Made by Freedomform 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: Wet-Ink is an interactive page that shows a decoding method for diffusion language models. A front moves through the text left to right and opens one token per step. A window of recent tokens behind it stays open to edits, and past that window the ink dries and tokens are frozen for good. The method was evaluated on LLaDA2.2-mini.

Who it suits: Researchers and students who work with diffusion language models and want to see how the method behaves. It also suits readers of the Wet-Ink paper who want an animation next to the text.

Industry | Who uses it | How they use it | Example
Research | Language model researchers | Watch the decoding loop and the window effect | Change W and see where revisions are still possible
Education | ML students | Learn how a sliding window limits edits | Play the animation from the start
Writing about AI | Technical writers | Explain the method with a live picture | Capture the animation at different window sizes

If you do not work with diffusion language models, you can skip it.

Where to use it: The page works in these places:

• A web browser: open the Space page on Hugging Face and play with it there.
• Windows, macOS or Linux: download the files and open index.html in a browser.

Use case | Who | What they get back
Watch the decoding loop | Researchers | An animation of tokens opening, staying editable and freezing
Tune the window | Students | A view of revision chances appearing and vanishing as W changes
Read the paper | Readers | A visual companion to the paper and its model card

How to set it up: Using the page online needs nothing. For a local copy you need a few things:

• Python with pip, to fetch the files
• The huggingface_hub package, which the first command installs
• A web browser

The steps download the Space and then open its main page by hand.

1. Install the Hugging Face package.
2. Download the Space into a folder called wet-ink.
3. Move into that folder.
4. Open index.html in your browser.

pip install -U huggingface_hub
hf download Freedomform/wet-ink --repo-type space --local-dir wet-ink
cd wet-ink

How to use it: Open the page and play the animation, then change the window size W and watch what happens to the tokens behind the front. The Space has no server part, only index.html and style.css.

Tool | What it does
W setting | Sets how many recent tokens stay editable
Animation | Plays the decoding loop on the page
hf download | Copies the Space's files into a folder on your computer

Example: You open the page and raise W. The animation then shows more tokens behind the front that can still be revised before they freeze. Lowering W shortens that stretch, so tokens freeze sooner.

For newcomers and experts: If you are new to this, think of the page as a picture of text being written from left to right. Open it, press play and move W once. You do not need any background to see the effect.

If you know the field, each step opens a single token at the front. The W tokens behind it can still change, and a T2T gate trained with the model controls those changes. Past W, a token is frozen. The Space names the Wet-Ink v5 model, LLaDA2.2-mini and Salesforce's ctrl. The paper site comes in English, Russian and Uzbek.

A front opens one token per step
A window of W tokens stays editable
The T2T gate decides on revisions
At distance W the tokens freeze

Good to know: The page is an animation, so it teaches the idea but does not run a model for you.

• License: mit, as the Space lists it.
• Cost to run: none named.
• Data it reaches: none named. A local copy downloads the Space's files from Hugging Face.
• Not for: generating text. It animates the loop and does not run a language model.

On Hugging Face, 13 November 2026

• Likes: 1
• Built with: static
• License: mit

Links

• Space: https://huggingface.co/spaces/Freedomform/wet-ink
• Files: https://huggingface.co/spaces/Freedomform/wet-ink/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.