---
title: Wavlm Phonemizer Word Detection
emoji: 🚀
colorFrom: green
colorTo: gray
sdk: gradio
python_version: 3.12.3
sdk_version: 5.47.0
app_file: app.py
pinned: false
license: apache-2.0
short_description: A simple space for word detection based on phonemes
---

# WavLM Demo

Some simple utility script to show how WavLM works and how to use it.
It is all based on WavLM Base + Phonemizer FR-IT

## Install

You need Python 3.12.

Use either uv (recommended), or pip.

```shell
# Using uv (recommended)
uv sync

# With pip
pip install -r requirements.txt`.
```

## main.py

This is the principal entry point.
Upon running, it will either capture audio from the microphone, or from a sample file.
Then, it will run an animate an inference, layer by layer and time step by time step.

![Animation of the inference](images/inference_animation.gif)

## app.py - Gradio interface

The Gradio interface is a web interface to demonstrate the word classification capabilities.

![View of the Gradio interface](images/gradio_space.png)

To launch Gradio, just run:

```shell
python app.py
```

And click on the web link!

### Advanced settings

The advanced settings will let give you fine-grained control over the generated answer.

It allows to circumvent shortcomings that a simple Arg Max decoder would give you.

![Alignment matrix](images/alignment_chart.webp)