PixAI Tagger v1.0

PixAI Tagger v1.0

Adds descriptive tags to an anime image in your browser, sorted into six categories with adjustable thresholds. Suits artists, dataset curators and anyone organizing anime art.

Made by pixai-labs 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: This Space is a demo of PixAI Tagger v1.0, a model that writes tags for anime images. You add a picture, press the tag button, and get tags grouped into general, character, style, copyright or series, meta and rating. You can copy the combined list or browse each group.

Who it suits: It suits artists, collectors and dataset curators who need consistent tags for anime pictures, and developers who want to see the model run before using it.

Industry | Who uses it | How they use it | Example
Digital art | Illustrators | Tag finished pieces for posting or archiving | Get a tag list for a character drawing
Dataset work | Curators | Label many images with a shared vocabulary | Tag a picture and copy the combined tags
Communities | Moderators and archivists | Sort images by character and series | Browse the character and series groups
Software | Developers | Read a handler around a six-category pipeline | Open handler.py in the repository

You can skip it if your images are not anime, since the model is built for anime art.

Where to use it:

• A web browser: open the Space page on Hugging Face
• Windows, macOS or Linux: run a copy with Python 3.11
• CPU only: the demo runs on a CPU, so no graphics card is needed

Use case | Who | What they get back
Tag one image | Artists | Combined tags plus six category lists
Tune the thresholds | Curators | Fewer or more tags per category
Add confidence scores | Developers | Tags with a score beside each

How to set it up: The online page needs nothing. A local copy needs Python 3.11 and access to the model.

• A web browser for the online demo
• Python 3.11 for a local copy
• The packages listed in requirements.txt
• A Hugging Face token with read access to the model, set as HF_TOKEN, when a Space loads it
• Or a local model folder, which skips the download

1. To use it online, open the Space page on Hugging Face.
2. To run a copy, install the Hugging Face hub package.
3. Download the Space into a local folder.
4. Move into the folder and install Gradio 6.14.0 and the listed packages.
5. Start the app.

pip install -U huggingface_hub
hf download pixai-labs/pixai-tagger-demo --repo-type space --local-dir pixai-tagger-demo
cd pixai-tagger-demo
pip install gradio==6.14.0
pip install -r requirements.txt
python app.py

To use a model folder you already have, the README runs it like this.

pip install -r requirements.txt
MODEL_DIR=/path/to/pixai-tagger-v1.0 python app.py

How to use it: Upload or paste an anime picture and press the tag button. Copy the combined tags, or look through the six category lists. Open the tag settings to change thresholds, pick the top tags, keep underscores or show confidence scores.

Example: You paste a character drawing and press the tag button. The page returns a combined list and six category lists. You raise the character threshold to see fewer, more certain tags, then copy the result.

For newcomers and experts: If you are new to this, tag one picture with the default settings and read the lists. Then open the settings and move one threshold to see how the tag count changes.

If you know the field, the thresholds default to the per-category settings that gave the best macro F1 in the full-vocabulary evaluation. The handler wraps the model's six-category Transformers pipeline. Inference uses the CPU at 32-bit float precision and serves a single request at once, with 2 torch threads unless you set TORCH_NUM_THREADS.

You upload or paste an anime image
The six-category tagger model runs on CPU
Thresholds filter tags in each category
You copy or browse the tags

Good to know: A few environment variables let you point the app at another model, a pinned revision or a local folder. Your HF_TOKEN value stays on the server side, out of the page and the API.

Title: Default threshold per category
General: 0.17 threshold
Character: 0.27 threshold
Style: 0.15 threshold
Copyright: 0.24 threshold
Meta: 0.17 threshold
Rating: 0.41 threshold

• License: MIT
• Cost to run: a Hugging Face token with read access to the model
• Data it reaches: the images you give it, and model files downloaded to the Hub cache
• Not for: non-anime images or heavy parallel use, since one request runs at a time

On Hugging Face, 13 November 2026

• Likes: 112
• Built with: gradio
• License: mit

Links

• Space: https://huggingface.co/spaces/pixai-labs/pixai-tagger-demo
• Files: https://huggingface.co/spaces/pixai-labs/pixai-tagger-demo/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.