Create readme.txt
Browse files- readme.txt +49 -0
readme.txt
ADDED
@@ -0,0 +1,49 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
title: Accent Classifier
|
3 |
+
emoji: 🎙️
|
4 |
+
colorFrom: indigo
|
5 |
+
colorTo: cyan
|
6 |
+
sdk: gradio
|
7 |
+
sdk_version: "3.38.1"
|
8 |
+
app_file: app.py
|
9 |
+
pinned: false
|
10 |
+
---
|
11 |
+
|
12 |
+
# Accent Classifier 🎙️
|
13 |
+
|
14 |
+
This Gradio app downloads a public YouTube or Vimeo video, extracts its audio, and classifies the speaker (as a proxy for accent) using a Hugging Face model. It’s perfect for demonstrating how to hook up **yt-dlp**, **ffmpeg**, and a **wav2vec2** pipeline in one slick interface—no rocket science required (just a little AI eavesdropping).
|
15 |
+
|
16 |
+
---
|
17 |
+
|
18 |
+
## 🛠️ How It Works
|
19 |
+
|
20 |
+
1. **Input**: You paste a *public* YouTube or Vimeo video URL into the Gradio textbox.
|
21 |
+
2. **Download**: Under the hood, `yt-dlp` fetches the video’s best audio stream and saves it as `downloaded_video.mp4`.
|
22 |
+
3. **Extract Audio**: `ffmpeg` converts that MP4 into a WAV file (`extracted_audio.wav`) at 16 kHz, mono—exactly what speech models crave.
|
23 |
+
4. **Classify**: A Hugging Face `pipeline("audio-classification", model="superb/wav2vec2-base-superb-sid")` processes the WAV file and returns a speaker ID (used here as an accent proxy) plus confidence.
|
24 |
+
5. **Cleanup**: Temporary files are removed automatically so you don’t end up with a cluttered folder.
|
25 |
+
6. **Output**: The app displays the predicted speaker ID and confidence percentage.
|
26 |
+
|
27 |
+
---
|
28 |
+
|
29 |
+
## 📦 Requirements
|
30 |
+
|
31 |
+
- **Python 3.8+**
|
32 |
+
- **yt-dlp** (installed automatically by the script if missing)
|
33 |
+
- **ffmpeg** (must be installed on your system and accessible via your command line)
|
34 |
+
- **gradio** (for the web interface)
|
35 |
+
- **transformers** (Hugging Face library)
|
36 |
+
|
37 |
+
> ❗ **Note**: If `yt-dlp` is not already installed, the code will install it at runtime. However, you must have `ffmpeg` installed manually. On macOS you can use Homebrew (`brew install ffmpeg`); on Ubuntu/Debian, `sudo apt-get install ffmpeg`; on Windows, download from [ffmpeg.org](https://ffmpeg.org/) and add it to your PATH.
|
38 |
+
|
39 |
+
---
|
40 |
+
|
41 |
+
## 🚀 Installation
|
42 |
+
|
43 |
+
1. **Clone or Download** this repository (the one containing `app.py` and this `README.md`).
|
44 |
+
2. **(Optional)** Create and activate a virtual environment:
|
45 |
+
|
46 |
+
```bash
|
47 |
+
python3 -m venv venv
|
48 |
+
source venv/bin/activate # macOS/Linux
|
49 |
+
venv\Scripts\activate.bat # Windows
|