bbc-esq/ctranslate2-faster-whisper-transcriber

Record audio and save a transcription to your system's clipboard with ctranslate2 and faster-whisper.

Record and Transcribe Audio Using Ctranslate2!

Record audio and save a transcription to your system's clipboard with ctranslate2 and faster-whisper.

Prerequisites

1. Python 3.10 or 3.11.
2. Git
3. git-lfs

Obtain Repository

Download the latest release and unzip the files to your chosen directory.

Setup

Navigate to the folder containing ct2_main.py, open a command prompt, and create a virtual environment:

python -m venv .

.\Scripts\activate

‼️On Linux and MacOS run: source bin/activate

python -m pip3 install --upgrade pip

Install Platform-Specific Libraries

• Linux

sudo apt-get install python3-dev

sudo apt-get install portaudio19-dev

• MacOS

brew install portaudio

Install Dependencies

pip3 install -r requirements.txt

Install Faster-Whisper

pip3 install faster-whisper==0.10.1

Nvidia GPU support

Only systems with GPU acceleration need to follow this step:

Go here for instructions on how to get the "cublas" and "cudnn" libraries and place them in the directory.
Alternatively, you can go here.

Usage

Within the virtual environment run this command:

python ct2_main.py

‼️ If you receive an error something similar to this qt.qpa.plugin: Could not load the Qt platform plugin "xcb" please check here for a possible solution..

EXAMPLE COMMANDS

The first time using the program, click "Update Settings" button to download the model. After that, you can change the model and quantization (and device) by simply changing the settings and clicking "Update Settings" again.

Click start recording, speak, then stop recording. Then just use control + v or right click "paste" the transcription into wherever you want; for example, into the chat box for your LLM!

Remember, anytime you want to restart the program, make sure to activate the virtual environment first!

ctranslate2-faster-whisper-transcriber's Issues

Models don't always change after updating.

After changing models and clicking the update button (sweet feature btw), on MX Linux 23.2 it appears to stay on the previous model, which is especially obvious when switching between medium and small or base.

Restarting the program seems to solve the problem though, so good for now.

Love the app btw.

Paste output duplicates and has inconsistencies

Using MX Linux 23.2 the paste results seem to duplicate a random amount. See my test results below trying the medium.en, small.en, and base.en models:

Where does sunshine come from?
Where does sunshine come from?
Where does sunshine come from?
Where does sunshine come from?
Where does sunshine come from?
Where does sunshine come from?
Where does sunshine come from?

Tell me about the complexities of quantum physics.

Tell me about the complexities of quantum physics.

Tell me where you're at right now.
Tell me where you're at right now.
Tell me where you're at right now.
Tell me where you're at right now.
Tell me where you're at right now.
Tell me where you're at right now.
Tell me where you're at right now.
Tell me where you're at right now.
Tell me where you're at right now.
Tell me where you're at right now.
Tell me where you're at right now.
Tell me where you're at right now.

Hello, can I help you?
Hello, can I help you?
Hello, can I help you?

Where do blue lights go at night?
Where do blue lights go at night?
Where do blue lights go at night?
Where do blue lights go at night?
Where do blue lights go at night?
Where do blue lights go at night?
Where do blue lights go at night?
Where do blue lights go at night?
Where do blue lights go at night?
Where do blue lights go at night?
Where do blue lights go at night?
Where do blue lights go at night?
Where do blue lights go at night?
Where do blue lights go at night?
Where do blue lights go at night?
Where do blue lights go at night?
Where do blue lights go at night?
Where do blue lights go at night?
Where do blue lights go at night?
Where do blue lights go at night?
Where do blue lights go at night?
Where do blue lights go at night?

Tell me how long it takes to get to New York City. Tell me how long it takes to get to New York City. Tell me how long it takes to get to New York City. Tell me how long it takes to get to New York City. Tell me how long it takes to get to New York City. Tell me how long it takes to get to New York City. Tell me how long it takes to get to New York City. Tell me how long it takes to get to New York City. Tell me how long it takes to get to New York City. Tell me how long it takes to get to New York City. Tell me how long it takes to get to New York City. Tell me how long it takes to get to New York City. Tell me how long it takes to get to New York City. Tell me how long it takes to get to New York City. Tell me how long it takes to get to New York City. Tell me how long it takes to get to New York City. Tell me how long it takes to get to New York City. Tell me

What can we do for you today, mister?
What can we do for you today, mister?
What can we do for you today, mister?

Could not load the Qt platform plugin "xcb" in "/usr/lib/x86_64-linux-gnu/qt5/plugins/platforms" even though it was found.

I'm on PopOS based on Ubuntu 22.04. Been getting the below error after running 'python ct2_main.py', no matter what solution I try. Any idea how to resolve this? Google and ChatGPT have not been helpful? Thanks in advance.

Updated cpu quantizations in config.yaml to: ['float32', 'int8_float32', 'int8']
Updated cuda quantizations in config.yaml to: ['float32', 'int8_float32', 'float16', 'int8_float16', 'int8']
qt.qpa.plugin: Could not load the Qt platform plugin "xcb" in "/usr/lib/x86_64-linux-gnu/qt5/plugins/platforms" even though it was found.
This application failed to start because no Qt platform plugin could be initialized. Reinstalling the application may fix this problem.

Available platform plugins are: vnc, offscreen, wayland-egl, minimalegl, linuxfb, vkkhrdisplay, wayland, minimal, eglfs, xcb.

Aborted (core dumped)