rbmikehuber/podcast-library-rb

Initial Setup

Prepare python env:

conda create --name podcasters
conda activate podcasters
pip install -r requirements.txt

Setup GCP credentials:

gcloud auth login
gcloud auth application-default login

Running

Start the backend:

cd backend/
uvicorn main:app --reload

Start the frontend:

cd frontend/
npm ci
npm run dev

Adding a podcast to the app

1. Upload the podcast audio file to a GCP storage bucket.
2. Use src/main.py to transcribe the podcast into a JSON file (be sure to set the correct language_code!).
3. Put the podcast audio file into backend/resources/audio-files/.
4. Put the transcript JSON into backend/resources/transcripts.
5. Add the two files to the podcasts array in backend/main.py:
   • The podcast audio file with key audio_file.
   • The transcript JSON with key transcript.
   • Add a nice name with key name.

Addding the full transcript text

1. After you have already uploaded the audio file to the GCP bucket (step 1 above)
2. Create a "new transcription" in the GCP console from the audio file you just added: https://console.cloud.google.com/speech/transcriptions/list?authuser=0&project=redbull-hack23szg-2116
3. Wait until the processing is complete, then click on your new transcript
4. Click the "DOWNLOAD" button at the top of the transcription and download the transcript as text
5. Move this file to backend/resources/transcripts
6. Add this file to the podcasts list at the top of main.py with the key transcript_only:

podcasts = [
    {
        "name": "Airton Cozzolino interview",
        "audio_file": "resources/audio-files/audio-files_msw_airton.mp3",
        "transcript": "resources/transcripts/msw_airton.mp3.json",
        "transcript_only": "resources/transcripts/msw_airton.txt"
    }
]

Notes

Find the sample rate of an MP3:

$ ffprobe msw_airton.mp3 
[...]
Input #0, mp3, from 'msw_airton.mp3':
  Duration: 00:16:36.36, start: 0.000000, bitrate: 127 kb/s
  Stream #0:0: Audio: mp3, 44100 Hz, stereo, fltp, 128 kb/s