Make your cards talk- natural voice text-to-speech with KokoroTTS

Card Read Aloud reads your review cards out loud in a natural, gentle human-sounding voice. It's built for medical students using AnKing and other med decks, so it says "I V", not "Roman four".

🎙️ A real neural voice, free and open source

The natural voice is Kokoro, a free, open-source neural text-to-speech model (Apache 2.0). It runs on your own computer and works offline. Your cards never leave your machine.

• 🗣️ Gentle voices to choose from: Heart (warm), Nicole (soft, almost a whisper), a Heart + Nicole blend, Bella, Emma, Michael and more
• ⚡ Speaks live, with no pre-rendering. There's no waiting to generate audio and no sound files added to your collection, so syncing stays fast
• 🎚️ Adjustable speed and volume

⚠️ Heads-up: one-time ~500 MB download is REQUIRED FOR NATURAL VOICE. The natural voice isn't bundled. When you turn it on (Tools → Card Read Aloud → Settings → Download), the add-on downloads the voice model and a small private Python engine, about 500 MB. This happens once. It lives inside the add-on's own folder, never touches Anki's Python, and the Remove button deletes all of it.

Don't want the download? Choose Built-in voice to use your computer's own voices instead.

💊 Medical pronunciation built in

• Shorthand is read the way it's said on the wards. IV, PO, BID and PRN are spelled out; q8h becomes "every 8 hours"; 15 mg/kg becomes "15 milligrams per kilogram"
• Roman numerals that are numbers stay numbers. Type IV, factor VIII and CN VII are read as 4, 8 and 7
• Ions and labs: Na⁺, K⁺ and HCO₃⁻ become sodium, potassium and bicarbonate; mEq/L, mmHg and cells/mm³ are read in full
• Biochemistry: acetyl-CoA, HMG-CoA, NADPH, G6PD, F-1,6-BP, G3P, 2,3-BPG, α-KG, CoQ10, CD4⁺, IL-6, TNF-α and β-lactam all come out right
• 1,400+ exact pronunciations (natural voice) for drug, bug, anatomy and biochemistry names the voice would otherwise mispronounce, checked against a full AnKing deck: antibiotics, antivirals, cardiac and psych drugs, -mabs, virus families, the urea cycle, amino acids, lipid metabolism and the lysosomal storage diseases
• Joined words are read in their parts: neoantigen is "neo antigen", oncofetal is "onco fetal", autoantibody is "auto antibody", thromboembolism is "thrombo embolism"
• Abbreviations are spelled out with natural stress: IgM, CEA, ADCC, EGFR, SIADH and DNA as letters; CAR T as "car T", Treg as "T reg"
• Add your own words in Settings

🧩 Reads the card the way you study it

• Front is read when the card appears and the back when you flip. Press T to replay the current side or stop
• Cloze-aware: the front says "blank" for the hidden part (plus the hint, if any); the back reads just the answer, or the whole sentence if you prefer
• Skips the clutter: images, sound buttons, collapsed hint fields (Lecture Notes, Missed Questions…), the AnKing countdown timer and the tag bar
• Plays nicely with your card audio: it waits for the card's own sounds to finish and stops when you flip or answer

⚙️ Settings

Open Tools → Card Read Aloud → Settings, or the Config button in the Add-ons window.

ℹ️ Good to know

• Tested on Anki 26.9 on macOS (Apple Silicon). Windows and Linux should work too; please report any problems on the support page
• The natural voice downloads its pieces from PyPI (uv, kokoro-onnx, onnxruntime) and from the kokoro-onnx release on GitHub, with Hugging Face as a backup
• Works alongside theme and answer-sound add-ons

📝 What's new

1.8: fixes cards being read twice when the add-on was installed both from AnkiWeb and from a file.

1.7:

• Joined words are read in their parts: neo antigen, onco fetal, auto antibody, thrombo embolism
• IgM, IgG, DNA, CEA and other abbreviations are spelled out with natural stress
• Roman numerals on their own are numbers, never "Roman two"
• About 640 new exact pronunciations, checked against a full AnKing deck

_{Voice: Kokoro-82M by hexgrad (Apache 2.0), run with kokoro-onnx. Add-on built with Claude by Anthropic.}

Install

• From AnkiWeb: Tools → Add-ons → Get Add-ons, then paste the add-on's code.
• From this repo: download card_read_aloud.ankiaddon from the Releases page and double-click it, or copy the card_read_aloud folder into Anki's addons21 folder.

Changes

See CHANGELOG.md.

Report a problem

Open an issue. Include your Anki version, your operating system and, for natural voice problems, the file card_read_aloud/user_files/natural_voice/log.txt.