Artwork Personalization Demo

Artwork Personalization Demo

Shows how to pick which poster to display for each title using CLIP image embeddings instead of viewing history. A static browser demo inspired by a public Netflix engineering article, not affiliated with Netflix.

Made by avantiwhenever 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 browser demo personalizes which poster shows for a title. It ranks poster images with CLIP image embeddings, not with a record of who watched what. The idea follows Netflix's MAPS approach to multimodal asset personalization. The author states that it is not affiliated with Netflix.

Who it suits: It suits recommender-system engineers, data scientists and students who want to see content-based artwork selection at small scale. It also suits product teams who want to understand the idea before building their own.

Industry | Who uses it | How they use it | Example
Streaming | Recommender engineers | Study artwork selection without interaction history | Open the demo and look at how posters are ranked
Marketing | Creative teams | Understand why different viewers see different images | Compare the posters shown for one title
Data science | Analysts | Explore CLIP embeddings as a ranking signal | Read app.js and the data folder
Education | Students | Learn content-based personalization | Follow the demo from poster to ranking

Anyone who needs a production recommender with real user data can skip it, since this is a small demo with posters and metadata from TMDB.

Where to use it: The demo runs in two places.

• A web browser, on the page of the Space at Hugging Face
• Windows, macOS or Linux, as a local copy opened in a browser

Use case | Who | What they get back
Poster ranking | Engineers | A choice of poster for each title, based on image embeddings
Idea walkthrough | Product teams | A working picture of how artwork personalization behaves
Code reading | Developers | A static page whose script and data are easy to inspect

How to set it up: The online page needs nothing installed. A local copy needs the following.

• Python and pip, only to download the files
• A web browser
• Room for the repository, which holds 620 files including the posters

1. Install the Hugging Face hub package and download a copy of the Space.
2. Move into the new folder.
3. Open index.html in a web browser.

pip install -U huggingface_hub
hf download avantiwhenever/netflix-personalization --repo-type space --local-dir netflix-personalization
cd netflix-personalization

How to use it: Open the page and browse the titles. Look at which poster the demo picks for each one. The page runs from index.html with a script, a style sheet, a data folder and a posters folder.

Tool | What it does
index.html | The page you open
app.js | The script behind the demo
data folder | Holds the demo's data
posters folder | Holds the poster images

Example: You open the demo and pick a title. The page shows the poster it ranks first for that title, chosen by comparing CLIP image embeddings.

For newcomers and experts: If you are new to this, think of a streaming service showing different cover images to different people. Open the demo, choose a title, and see which poster the ranking puts first.

If you know the field, the approach scores posters from the images themselves through CLIP embeddings, which removes any reliance on ID-keyed interaction logs. The author's full source, data pipeline and documentation sit in a separate repository, including an offline validation of the ranking against popularity and random baselines. This Space holds the browser demo only.

Posters and metadata from TMDB
CLIP image embeddings
Ranking picks a poster per title
The demo shows the chosen poster

Good to know: The Space is a small teaching demo. The detailed validation lives outside it, so do not treat the demo as proof of the method.

• License: MIT
• Cost to run: none named
• Data it reaches: posters and metadata from TMDB, shipped in the repository
• Not for: claiming any link to Netflix, which the author says there is none

On Hugging Face, 13 November 2026

• Likes: 2
• Built with: static
• License: mit

Links

• Space: https://huggingface.co/spaces/avantiwhenever/netflix-personalization
• Files: https://huggingface.co/spaces/avantiwhenever/netflix-personalization/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.