podcast host.
use speech transcription to semantic search a DB of quotes in real time. then use an llm to highlight the relevant parts of quotes that fit the transcription
playlist of videos created with quo-host
Start poetry shell with poetry shell
- Take edited audio file and transcribe it to text with
transcriber/transcribe_assemblyai.py
by changingFILE_URL
- The transcribed json will be stored as
transcript.json
- Run the AI process with
transcriber/simulate.py
withsimulate_transcription_process_from_json('transcript.json')
- AI processed result will be stored in
logs/
folder 4a. Get missing covers withtranscriber/extract_covers.py
and editLOG_FILE
- in
public/index.html
changelog_file_path
timestamp_file_path
andaudio_file_path
(audio needs to also be inpublic/
folder) - check for missing covers in
transcriber/covers.json
- start the python server with
python server.py
- start the node server with
npm start
- open
localhost:3000
in your browser - click on the play button to start the audio and see the highlights (3s delay w/ stopgap solution)
Run OBS in 1080p and use the browser source to show the highlights on the screen:
/Applications/Google\ Chrome.app/Contents/MacOS/Google\ Chrome --app=http://localhost:3000
then in menu bar Chrome > services > resize1080p
ctrl-c
to stop the servers- get the chapters from
transcriber/transcript_to_chapters.py
and editTRANSCRIPT_PATH