Whisper GUI

An interface to streamline transcriptions using Whisper.

• transcribe video files up to ~ 25 MiB
• transcribe audio files of any size
  (Audio files larger than 25 MiB are split into segments with an overlap of 2 seconds. Finally the segment transcriptions are merged.)

Requirements

• whisper (see requirements online)
• nicegui
• pywebview
• openpyxl
• pydub
• simpleaudio

Developer

Notes

🐍 Python Version
As of now (2024/10/19) Whisper works with Python 3.9 or 3.10. Make sure to select correct version in your IDE (note to self: in VS Code select Python 3.9.0 ('.venv') in lower right corner).

ℹ️ Virtual Environment
Its best practice to setup a virtual environment for the project. Make sure to activate it (.\.venv\Scripts\activate) before installing libraries!

Setup Example

1. Create a virtual environment: .venv
   Make sure to initialize the virtual environment using the correct python version (3.9 or 3.10 see notes):

   C:\Users\\[UserName]\AppData\Local\Programs\Python\Python39\python.exe -m venv .venv
   

2. Activate the virtual environment

   .venv\Scripts\activate
   

3. Install the following dependencies

   pip install git+https://github.com/openai/whisper.git  
   pip install blobfile  
   pip install nicegui  
   pip install pywebview
   pip install openpyxl
   pip install pydub
   pip install simpleaudio
   

4. To use the GUI on your machine, make sure pre-requisites (mainly NVIDIA CUDA and FFMPEG) are installed. See openai/whisper#1463 for more info.

Build Executable

0. first time, install pyinstaller in python environment

   (.venv) pip install pyinstaller
   

1. run nicegui-pack

   (.venv) nicegui-pack --windowed --name "Whisper GUI" main.py
   

2. copy the following files and folders to .\dist\Whisper GUI\_internal

   • sound_effect_finished.wav
   • .\.venv\Lib\site-packages\whisper
   • .\.venv\Lib\site-packages\tiktoken_ext

Whisper Offline Use

Add folder C:\Users\[UserName]\.cache\whisper. Move the files docs/vocab.bpe and docs/encoder.json to this folder. Update your local copy of openai_public.py. If you created a venv this file is located in .venv\Lib\site-packages\tiktoken_ext\openai_public.py otherwise it probably is in C:\Users\[UserName]\AppData\Local\Programs\Python\Python310-32\Lib\site-packagespython3.9\site-packages\tiktoken_ext\openai_public.py. Remove the URL "https://openaipublic.blob.core.windows.net/gpt-2/encodings/main/" and replace it with your local copy, e.g.:

    def gpt2():
    mergeable_ranks = data_gym_to_mergeable_bpe_ranks(  
        vocab_bpe_file="C:/Users/[Username]/.cache/whisper/vocab.bpe",   
        encoder_json_file="C:/Users/[Username]/.cache/whisper/encoder.json",  
    )