Julien Fastré
6b561bdda1
This update introduces a table to the README that outlines the accuracy, speed, and best use cases for Whisper models in French. It provides clarity for users selecting a model for French transcription tasks.
84 lines
2.7 KiB
Markdown
84 lines
2.7 KiB
Markdown
# Project README
|
|
|
|
## Setting Up the Project Environment
|
|
|
|
Follow the steps below to set up your project environment:
|
|
|
|
### 1. Create a Virtual Environment
|
|
|
|
To isolate the project dependencies and prevent conflicts, create a virtual environment using the following commands:
|
|
|
|
```bash
|
|
# Create the virtual environment
|
|
python3 -m venv venv
|
|
|
|
# Activate the virtual environment:
|
|
source venv/bin/activate
|
|
```
|
|
|
|
### 2. Install Project Dependencies
|
|
|
|
Once the virtual environment is activated, install the required dependencies using `pip`:
|
|
|
|
```bash
|
|
pip install -r requirements.txt
|
|
```
|
|
|
|
Make sure you have a `requirements.txt` file in the project directory with the list of all required dependencies.
|
|
|
|
## Running the Script
|
|
|
|
To run the Python script `transcribe.py`, you need to provide an audio file as a parameter. Use the following command:
|
|
|
|
```bash
|
|
python transcribe.py <audio_file>
|
|
```
|
|
|
|
- Replace `<audio_file>` with the path to your audio file.
|
|
- Example:
|
|
|
|
```bash
|
|
python transcribe.py sample_audio.wav
|
|
```
|
|
|
|
## Where is downloaded the AI Whisper model ?
|
|
|
|
When using the `openai-whisper` package, the AI Whisper model is downloaded and stored in a local cache directory. By
|
|
default, it is stored under the user's home directory in the following path:
|
|
|
|
```plaintext
|
|
~/.cache/whisper/
|
|
```
|
|
|
|
Here:
|
|
|
|
- `~` refers to the user's home directory.
|
|
- `.cache/whisper/` is the folder where the models are cached.
|
|
|
|
The cache directory contains the downloaded model files, which are reused in subsequent runs to avoid re-downloading
|
|
them. Specifically:
|
|
|
|
- Whisper downloads the model files when they are first used, based on the requested model size (e.g., `base`, `medium`,
|
|
or `large`).
|
|
|
|
If you need to modify or relocate the cache directory, you can set the `WHISPER_CACHE_DIR` environment variable to
|
|
specify a custom path for storing these files.
|
|
|
|
## Which whisper model for French ?
|
|
|
|
| Model | Multilingual? | Accuracy in French | Speed | Best For |
|
|
|-----------|--------------|--------------------|-------|----------|
|
|
| **tiny** | ✅ Yes | ❌ Poor | 🚀 Fastest | Basic transcription, very simple audio |
|
|
| **base** | ✅ Yes | ⚠️ Okay | 🔥 Fast | Short/simple French audio |
|
|
| **small** | ✅ Yes | ✅ Good | ⚡ Fast | General French transcription |
|
|
| **medium**| ✅ Yes | ✅✅ Very Good | ⏳ Slower | Better accuracy in noisy audio |
|
|
| **large** | ✅ Yes | ✅✅✅ Best | 🐢 Slowest | High accuracy, complex speech |
|
|
|
|
|
|
## Notes
|
|
|
|
- Ensure that your virtual environment is activated before running the script.
|
|
- If you encounter any missing dependencies, double-check your `requirements.txt` file and re-run the installation
|
|
command.
|
|
|
|
Happy coding! |