Thematic Transcription

A tool to transcribe video to audio using Whisper API for thematic analysis.

Attention weights to capture dependencies and interactions between words in the audio.

The following is split up into a few sections: (1) Focused on transcription and performing speaker diarization, meaning separating speakers in the audio (2) Thematic analysis of the transcription (in progress)

Installation

1. Set up Whisper API GitHub instructions

If you do not have admin access, make sure to run as administrator when installing

2. Install dependencies in your command prompt

pip install whisper pip install python-docx pip install fpdf

3. Add audio file to the same directory or add the correct path in code

4. Modify code as necessary such as changing the name of the audio file, language, etc.

5. Run the script. Replace your_script.py with the name of your script

python your_script.py

6. See saved transcription as a word (.docx) and pdf (.pdf) file in the same directory as the audio file

Acceptable file types

The acceptable file types¹ are:

• m4a: MPEG-4 Audio File
• mp3: MPEG-1 Audio Layer 3 File
• webm: WebM Audio/Video File
• mp4: MPEG-4 Video File
• mpga: MPEG Audio File
• wav: Waveform Audio File
• mpeg: MPEG Movie File

Links to audio files are not supported at the moment.

Requests per minute

50 requests per minute¹

File size

Up to 25MB¹

Command prompts

Transcribe audio

1. cd into the directory where the audio file is located
2. whisper transcribe --filename <filename> --language <language> --output <output_filename>
   • filename is the name of the file you want to transcribe
   • language is the language of the audio file
   • output is the name of the file you want to save the transcription as

Transcribe audio with speaker separation

1. cd into the directory where the audio file is located
2. whisper transcribe --filename <filename> --language <language> --output <output_filename> --speaker-separation
   • filename is the name of the file you want to transcribe
   • language is the language of the audio file
   • output is the name of the file you want to save the transcription as

Transcribe audio with speaker separation and speaker labels

1. cd into the directory where the audio file is located
2. whisper transcribe --filename <filename> --language <language> --output <output_filename> --speaker-separation --speaker-labels
   • filename is the name of the file you want to transcribe
   • language is the language of the audio file
   • output is the name of the file you want to save the transcription as

Transcribe audio with speaker separation and speaker labels and punctuation

1. cd into the directory where the audio file is located
2. whisper transcribe --filename <filename> --language <language> --output <output_filename> --speaker-separation --speaker-labels --punctuation
   • filename is the name of the file you want to transcribe
   • language is the language of the audio file
   • output is the name of the file you want to save the transcription as

Transcribe audio with speaker separation and speaker labels and punctuation and profanity filter

1. cd into the directory where the audio file is located
2. whisper transcribe --filename <filename> --language <language> --output <output_filename> --speaker-separation --speaker-labels --punctuation --profanity-filter
   • filename is the name of the file you want to transcribe

License

MIT

Footnotes

1. Whisper API FAQ ↩ ↩² ↩³