whisper-to-cards

Whisper-to-Cards

Turn audio (talks, lectures, tutorials) into accessible notes + tagged, actionable flashcards (Anki). Works in GitHub Codespaces and publishes a public /demo/ page via GitHub Pages.

Live demo: https://tkhatton.github.io/whisper-to-cards/
Repo: https://github.com/TKHatton/whisper-to-cards

Quick Start (TL;DR)

1. Put audio in inputs/ (mp3/m4a/wav).
2. Process one file:

   scripts/run_one.sh "inputs/YourFile.mp3" "Odin :: Foundations" small
   

3. Make a deck (pick one):

poetry run python scripts/make_decks.py --mode ds
# or
poetry run python scripts/make_decks.py --mode git

1. (Optional) Publish demo:

scripts/publish.sh

One-Time Setup

These are already added in this repo. If anything’s missing, re-create from the chat.

chmod +x scripts/run_one.sh
chmod +x scripts/make_decks.py
# (optional) if you have publish.sh and auto_run.py:
chmod +x scripts/publish.sh

For .apkg export (optional): poetry add genanki

What the scripts do

scripts/run_one.sh <audio> "<Deck Name>" <model>

Runs ASR → segment → structure → render → bundle.

• model: small (fast, default) or medium (more accurate).

scripts/make_decks.py --mode ds|git

Builds tagged Anki decks from your run:

• --mode ds: Data Structures (definitions, when-to-use, safe Big-O).

• --mode git: Git/VS Code (how-to, commands, concepts).

• Outputs: outputs/*.csv and (if genanki installed) outputs/*.apkg.

• scripts/publish.sh (optional) Copies your featured deck to outputs/deck.apkg, bundles outputs/, and pushes so /demo/ updates.

Automatic Mode (Drop-a-file → Runs)

Leave this watcher running in one terminal tab:

poetry run python scripts/auto_run.py

Now whenever you add/modify an audio file in inputs/, it will:

• run run_one.sh on that file,

• generate a DS deck by default (change to git inside auto_run.py),

• (optionally) publish (uncomment the line inside auto_run.py).

  The watcher stays idle when no new files appear. Stop with Ctrl+C.

Demo Site (GitHub Pages)

Local/Codespaces: open the forwarded port URL → `/demo/`.

Public: `https://tkhatton.github.io/whisper-to-cards/demo/`

The demo iframe reads `outputs/notes.html`; buttons download PDF/CSV/APKG/ZIP.

Troubleshooting

• Script says “Usage: … path/to/audio.mp3” → you didn’t pass a real file path.

• Nothing happens in watcher → add/modify a file in inputs/.

• 404 in demo iframe → ensure outputs/notes.html is committed & Pages is deployed.

• Windows line endings → sed -i 's/\r$//' scripts/*.sh

• Force add generated files (ignored by .gitignore): git add -f outputs/ dist/ —

This site is open source. Improve this page.