DuET Piano to Score

DuET Piano to Score

Transcribes piano recordings into readable musical scores, from an uploaded file or a bilibili or 抖音 link. Pianists and teachers get a playable score with MusicXML and PDF downloads.

Made by Wayne42 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: DuET Piano to Score listens to a piano recording and writes it out as a readable musical score. You give it an audio or video file, or a link to a video on bilibili or 抖音, and it transcribes the first 240 seconds. The result can be played back in the page or downloaded as MusicXML or PDF.

Who it suits: Pianists, teachers and students who want a score from a recording, and developers who want to see a piano transcription demo.

Industry | Who uses it | How they use it | Example
Music education | Piano teachers | Get a readable score from a student's recording | Turning a short practice clip into a score to mark up
Music making | Pianists and arrangers | Draft sheet music from a performance | Exporting MusicXML to edit in a notation program
Research | Audio researchers | Try a piano transcription model on real recordings | Running a clip and reading the report it gives

You can skip it if you need a whole long recording transcribed, since the online demo stops after 240 seconds.

Where to use it:

• A web browser, on the Space's own page at Hugging Face
• A copy on your own computer, using Python 3.12 on Windows, macOS or Linux

Use case | Who | What they get back
Transcribe a recording | Pianists and teachers | A score with playback, a cursor that follows the notes and automatic scrolling
Export a score | Arrangers | MusicXML and a PDF of the same score

How to set it up:

• A browser, if you only use the Space online
• Python 3.12 and pip, if you run a copy yourself
• The system packages in packages.txt, installed with apt-get
• A TikHub key named TIKHUB_API_KEY, kept as a Space secret, only for the bilibili and 抖音 link inputs

1. Open the Space's page on Hugging Face. Nothing needs installing for audio or video upload.
2. For a local copy, install the huggingface_hub package and download the Space's files.
3. Move into the new folder, then install Gradio and the packages in requirements.txt.
4. Start the app with python and the path to its main file.

pip install -U huggingface_hub
hf download Wayne42/DuET-demo --repo-type space --local-dir DuET-demo
cd DuET-demo
pip install gradio==6.30.0
pip install -r requirements.txt

How to use it: Choose an input source on the page: Upload audio/video, bilibili or 抖音. Upload a file or paste a link, then run the transcription. On your own computer, start the app first and open the page it serves.

python demo/app.py

Tool | What it does
Upload audio/video | Takes a file from your computer
bilibili | Takes a video link, a short link or a link with a part number
抖音 | Takes a normal video link, a short link or the full share text
Web score | Plays the score with a following cursor and automatic scrolling

Example: You pick Upload audio/video and add a short piano recording. The page transcribes the first 240 seconds and shows a score you can play, then you download it as MusicXML or PDF.

For newcomers and experts: If you are new to this, upload a short piano clip and watch the score appear. Play it back to hear whether the notes match your recording.

For experts, the model is DuET, a dual-encoder transformer for piano transcription, and the demo is version 1.1.0. MusicXML comes from engraving 1.1.0 with contextual key signatures and note spelling. The PDF is a MuseScore rendering, so browser pagination can differ. Only the demo is capped at 240 seconds, not the model.

Choose upload, bilibili or 抖音
Audio is fetched or read
DuET transcribes the piano audio
The score is engraved as MusicXML
You play it or download MusicXML or PDF

Good to know:

• Built with Gradio 6.30.0 and Python 3.12, and the main file is demo/app.py
• License: mit
• Cost to run: the link inputs call TikHub and need a TikHub key, and no other key or paid service is named
• Data it reaches: your uploaded audio, TikHub, bilibili and 抖音 for links, and a private audio cache kept for one hour
• Not for: YouTube links, which the demo does not offer, or accounts that need a login, since it asks for no cookies or login
• Only audio is exported from a video, and no video file is saved

On Hugging Face, 13 November 2026

• Likes: 10
• Built with: gradio
• License: mit

Links

• Space: https://huggingface.co/spaces/Wayne42/DuET-demo
• Files: https://huggingface.co/spaces/Wayne42/DuET-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.