Answer your cards out loud. Voice Review listens, checks every idea you said, and suggests the button.
Built for medical school (Step 1 / CBSE style cards), but it works on any deck. You're graded on the ideas, not the exact wording: "middle ear and sinus infections" counts for "otitis media, bacterial sinusitis", and Whisper's "Vile Felix" still counts as Weil-Felix.
How it works
- A card appears. Say your answer. Pauses mid-thought are fine.
- Say "done" (or press Space, or set it to end after a pause you choose).
- The answer is shown with each idea ticked ✓ or crossed ✗, a short high-yield hook, and the suggested button.
- Say "next" or press Enter to accept, or say/press "again", "hard", "good", "easy" to choose yourself. Say "why" for a one-line explanation of what you missed.
Grading you can predict
Each card is split once into its key ideas ("core" and "detail"). Every idea gets its own yes/no check, and plain rules turn those into a button. The rules are shown in Settings in plain English. Presets: Lenient, Board (default) and Strict, or set your own thresholds.
• Again: under half of the core ideas, or you said something wrong
• Hard: at least half of the core ideas, but not all
• Good: every core idea
• Easy: everything, quickly
When it isn't sure, it doesn't guess: it offers you both buttons.
In testing on 94 spoken-style answers to real Step 1 cards, it picked the right button 97% of the time, asked you in 2%, was never too harsh, and caught 60 of 60 wrong statements with no false alarms.
Everything is a setting
• Speaking: silent, read only what you missed, read the feedback, or read the whole card
• Listening: starts automatically, or only when you press a key
• Ending an answer: say "done", a pause of your chosen length, or a key / remote button
• Moving on: by voice, by key, or automatically after a correct answer
• Stop button (Shift+X) turns the mic fully off when you take a break
• Works with remotes and game controllers (one-click setup for Contanki)
Keys: L listen · Space end answer · Enter accept · ? why · Shift+R repeat · Shift+S speaking mode · Shift+M mic mode · Shift+K skip · Shift+X stop/resume. All can be changed.
Setup (about 5 minutes)
Open Tools → Voice Review → Set up API keys… Each key has a "How do I get this?" button with step-by-step instructions, and a Test button.
• Groq (required, free tier): speech-to-text (Whisper) and explanations. The free tier covers a few hundred cards a day.
• TypeSafe (required): the fast grader (Jev). Roughly $0.30–0.60 a month at a few hundred cards a day. New accounts may go through a short waitlist.
• Google Cloud (optional): a more natural voice. Without it, your computer's built-in voice is used (Mac, Windows or Linux).
Privacy
• Audio is sent only while you're speaking. Silence never leaves your computer.
• Your keys are stored only on your computer, in the add-on's folder, and are never synced or shared.
• Card text is sent to Groq (to find each card's ideas, once per card) and to TypeSafe (to grade).
Requirements: Anki 25.09 or newer, a microphone, and an internet connection.
Tips
• Use Tools → Voice Review → Prepare cards before a session so the first cards don't wait.
• If it ever grades you unfairly, pick the button you think is right. Your choice always wins.
• Problems? Turn on the debug log in Settings and include it in your report.