WhisperGPTTranscriber is a Python-based tool that efficiently converts audio files to text using OpenAI's Whisper model and enhances the transcription accuracy with GPT-4 post-processing. It supports large audio files, offers conditional GPT-4 enhancement, and allows customizable output locations.
- Audio Transcription: Utilizes the Whisper model for efficient audio-to-text conversion.
- GPT-4 Post-Processing: Offers optional post-processing with GPT-4 for enhanced transcription accuracy or a direct summary of key facts.
- Large File Support: Capable of processing audio files larger than the standard size limits of 25 MBs.
- Customizable Output: Allows users to specify output file locations.
- Install openai and pydub (mp3 handling):
pip install openai pydub - Add the openai key to the .py file
- Change the GPT Model if you want. I am using "gpt-4-1106-preview"
The script takes a while (minutes). Especially when the mp3 is long. As a progress Statement, it will print the transcripts of the "chopped" mp3 into the terminal.
Example: Download any Youtube Video as .mp3. There are a billion converters out there. Like https://notube.cc/de/youtube-app-v103 Then...
This will run the script and save the transcript in the same folder as the .py file:
python3 whisper.py path/to/audio.mp3
This will run the script, post-process it and then save the transcript in a designated folder:
python3 whisper.py path/to/audio.mp3 --gpt_post_process --output_file path/to/output.txt
The GPT post-processing feature in WhisperGPTTranscriber uses OpenAI's GPT-4 model to refine and enhance the transcriptions. After the initial transcription with Whisper, the text is passed through GPT-4, which corrects grammatical errors, clarifies ambiguous language, and ensures proper spelling, especially of specific terms or names.
- Technical Meetings: Improves the accuracy of technical jargon in transcriptions from IT or scientific discussions.
- Educational Lectures: Enhances clarity in transcribed lectures, ensuring correct terminology and coherence.
- Business Conferences: Polishes transcripts of business conferences, focusing on correct names of companies, products, and industry-specific terms.
- Medical Dictations: Ensures medical terms and drug names are accurately transcribed from doctors’ audio notes.
- Legal Proceedings: Refines transcriptions of legal proceedings, where accuracy of names, laws, and legal terminology is crucial.