Lecture Search — Semantic Card Finder & AI Gap Analysis
Lecture Search bridges the gap between your lecture slides and your existing Anki collection. It uses local semantic embeddings (SentenceTransformers with Apple Silicon GPU / CUDA acceleration) to search hundreds of thousands of flashcards in seconds, finding cards matching your PDF or PPTX slides.
If your deck is missing key lecture concepts, the built-in AI Gap Analysis powered by Google Gemini automatically identifies uncovered topics and generates high-yield Cloze flashcards with embedded slide images.
✨ Key Features
-
Instant Semantic Search:
- Searches your entire Anki collection (150,000+ cards) in seconds using local AI embeddings (
all-MiniLM-L6-v2).
- Supports both PDF and PPTX lecture presentations.
- Granular matching: choose Per Slide or Whole Document granularity.
-
One-Click Deck Movement:
- Select matched cards and move them directly into a new target lecture deck (e.g.,
Lecture::Immunology Ch4).
-
AI Gap Analysis (Gemini Integration):
- Compares your lecture slides against your existing cards.
- Finds high-yield gaps and generates cloze-deletion cards with embedded slide images in the
Extra field.
-
100% Local & Fast:
- Semantic embeddings run entirely on your local machine with Apple Silicon (MPS) and NVIDIA (CUDA) GPU acceleration.
Step 3 — Open Lecture Search
After the install completes, restart Anki and go to Tools → Lecture Search... (or press Ctrl+Shift+L / Cmd+Shift+L).
⚙️ Configuration
Open Tools → Add-ons → Lecture Search → Config:
similarity_threshold (default: 0.4): Cosine similarity cutoff (0.0 to 1.0).
default_deck_prefix (default: "Lecture"): Default prefix for created decks.
gemini_api_key: Optional Gemini API key for AI Gap Analysis (get a free key at aistudio.google.com).
gemini_model (default: "gemini-2.5-flash"): Gemini model for card generation.
cards_per_gap (default: 2): Number of cloze cards generated per identified gap.
⌨️ Shortcut
- Mac / Windows / Linux: Ctrl+Shift+L (or Cmd+Shift+L on macOS) opens Lecture Search from anywhere in Anki.