nathangathright/transcripts-to-json

Transcripts to JSON

Whisper.cpp supports word-by-word timestamps, but doesn’t support exporting them in the Podcast Namespace JSON structure. This is a quick guide on how to do that.

Transcribing with Whisper.cpp

1. Setup whisper.cpp:

   # Clone whisper.cpp and navigate into it’s directory:
   git clone https://github.com/ggerganov/whisper.cpp.git && cd whisper.cpp/
   
   # Download a language model:
   bash ./models/download-ggml-model.sh base.en
   
   # Build the project:
   make

2. Transcribe an audio file with the --output-json and --max-len arguments:

   ./main -m ./models/ggml-base.en.bin -f <input.wav> -oj -ml 1
   
   # **Warning**
   # The main example currently runs only with 16-bit WAV files, so make sure to convert your input before running the tool. For example, you can use ffmpeg like this:
   ffmpeg -i input.mp3 -ar 16000 -ac 1 -c:a pcm_s16le input.wav

Using transcripts-to-json

1. Setup transcripts-to-json:

   # Clone transcripts-to-json and navigate into it’s directory:
   git clone https://github.com/nathangathright/transcripts-to-json.git && cd transcripts-to-json/

2. Convert

   # SRT to JSON
   python srt-to-json.py <input.srt>
   
   # JSON to JSON
   python json-to-json.py <input.json>