AnkiWeb addon 1905058369
Flips an existing bilingual deck’s direction and generates offline pronunciation audio for the newly fronted language using local language and text-to-speech models.
Open on AnkiWeb GitHub Find alternatives
active
| Min Anki | Max Anki | Updated |
|---|---|---|
| 2.1.1 | 2.1.66+ | 2026-09-20 |
Loading…
Generates free text-to-speech audio with EdgeTTS and offline engines, saving files to cards or streaming during review, with batch generation, voice mapping rules and resumable checkpoints.
Generates contextual written explanations and spoken audio for target words in sentence cards using OpenAI models and selectable TTS engines, with one-click or batch processing.
Fixes broken Cambridge Dictionary audio links in existing Wordwise cards by scanning the collection and replacing them with embedded offline TTS audio.
Generates vocabulary flashcards through multiple LLM providers, adding definitions, pronunciations, text-to-speech, etymology, synonyms, antonyms, examples, Japanese support, and LaTeX-rendered math cards.
Generates Anki language-learning decks from word lists, adding translations, images, audio pronunciations, Tatoeba examples, and two-way cards, usable standalone or as a Windows-only addon.
Automatically fills French flashcard fields—translations, grammar, gender, plurals, pronunciation, and Google TTS audio—using an offline lexical database and two custom note types.
Takes an existing bilingual deck and flips its direction — for example, turns a Japanese-front/English-back deck into an English-front/Japanese-back deck — and generates natural pronunciation audio for the newly-fronted language. Everything runs locally and offline: a local language model reads the deck's real fields and writes the new card layout, and a local text-to-speech engine generates the audio. No cloud services, no accounts, no data ever leaves your machine.
Why this exists: I though about this idea because there are a lot of excellent, carefully-built decks for people learning Japanese( I am a Japanese learner) but not nearly as much equally well-structured material for people learning English. Rather than build a new deck from scratch, this addon takes a deck that already works and reverses it, so the same quality carries over to a much wider audience. (This is not just to be used with Japanese Decks!)
Before you install, please read:
How to use it:
Full documentation, source code, and license (MIT) at: github.com/kaisalayasa/Language-Switch-Anki-Add-on
An Anki addon that takes an existing bilingual deck and flips its direction — for example, turning a Japanese-front/English-back deck into an English-front/Japanese-back deck — and generates natural pronunciation audio for the newly-fronted language. Everything runs locally and offline: a local LLM reads the deck's real fields and writes the new card layout, and a local TTS engine generates the audio. No cloud services, no accounts, no data ever leaves your machine.
There are a lot of excellent, carefully-built decks out there for people learning Japanese — Core 2000 being a well-known example. But the reverse doesn't really exist: there isn't nearly as much equally well-structured material for people trying to learn English. Rather than build a new English-learning deck from scratch, it seemed simpler to just take a deck that already works — good example sentences, good structure, years of community refinement — and reverse it: flip which language is on the front, and generate real pronunciation audio for it. That way the quality of the original deck carries over, but now a much wider audience can actually use it.
TODO.md for the small list of known open issues..ankiaddon file — see Installing below if you're building it yourself).This addon's own code is MIT-licensed (see LICENSE) — free to use, modify, and
share.
Everything it downloads at runtime was individually checked for license compatibility:
| Component | License |
|---|---|
| Piper (TTS engine) | MIT |
| llama.cpp (LLM runtime) | MIT |
| Qwen2.5-7B-Instruct (language model) | Apache-2.0 |
Piper voice en_US-ljspeech-high |
Public domain |
Piper voice en_GB-alba-medium |
CC BY 4.0 (attribution: "Alba" voice via rhasspy/piper-voices) |
| ffmpeg (Windows/Linux build, used only to compress generated audio) | LGPL-2.1+ |
| ffmpeg (macOS build, same purpose) | GPL-3.0+ — invoked as a separate subprocess only, never linked into this addon's code, so its license doesn't apply to this addon's own code (the same way any app can shell out to git or ImageMagick without becoming GPL itself) |
None of these are bundled in this repository — the addon downloads each one from its own official source (GitHub releases, HuggingFace, evermeet.cx for the macOS ffmpeg build) the first time it's needed, and caches it locally.
The full technical write-up — architecture, design decisions, verified Anki/llama.cpp API
facts, and the reasoning behind every non-obvious choice in this codebase — lives in
CLAUDE.md and the docs/ folder. TODO.md tracks known
open issues.