Import from Anki: the honest guide to moving your deck
You have a deck. Possibly a large one, built over years, and you are not going to abandon it because a website told you to. This page is about moving it: what the steps actually are today, what comes across and what does not, and what happens to the cards once they are here.
Two things before anything else. The migration is manual and it has one genuinely annoying moment — you have to open the exported file in a text editor and delete two lines by hand. And your review history does not come with you; your cards do. Both are written out below rather than discovered by you halfway through.
What transfers, and what does not
| What | Comes across? | |
|---|---|---|
| Your cards — the text on both sides | Yes | All of it. That is what the file contains and what the importer reads |
| Audio and phonetics | Yes — rebuilt, not copied | Your Anki media files are not carried over. Instead every card gets a fresh natural voice on both sides and an IPA transcription, generated automatically in any of 76 languages. If you had audio in Anki you needed add-ons for it; here it is simply there, on every card, including the ones that never had it |
| Your scheduling history — intervals, ease, due dates | No | This is the real one. Scheduling lives in Anki's database, not in the export, so imported cards start their schedule fresh |
About that scheduling history. Losing it sounds worse than it is in practice. The scheduler adapts from your very first ratings, so a card you genuinely know gets pushed straight back out — you are re-confirming it, not re-learning it, and a mature deck re-settles quickly. It costs you a few sessions of easy reviews, not months.
But be honest with yourself about which one you are: if the value of your collection is the history itself — years of tuned intervals across twenty thousand mature cards — then that is the thing you would be giving up, and it is a perfectly good reason to stay where you are.
How to do it: four steps, one of them annoying
1. In Anki on a computer, go to File ▸ Export. Not AnkiWeb, not AnkiMobile, not AnkiDroid — the plain-text export lives in the desktop app only. That is the one hard requirement here.
2. Set Export format to Notes in Plain Text (.txt), pick your deck, and export. The format will not already be selected — Anki defaults to its own .apkg deck package, which we cannot read. Leave the five checkboxes alone; they are off by default and off is what you want.
3. Open the saved file and delete the two lines at the top that start with #. They will read #separator:tab and #html:false. This is the annoying step, and it is the only one — any text editor will do.
4. In Taalhammer, open the collection you want the cards in and click Import. Creator ▸ My Library ▸ your collection ▸ Import ▸ Choose CSV file from disk ▸ submit. Make sure the collection's language pair matches the two columns of your file, in that order — the importer takes the languages from the collection and assigns the first column to the source language and the second to the target. It does not look at the text to check. If your export has them the other way round, every card imports backwards, with no warning. Two ways to avoid that: reorder the columns before importing, or add one line at the very top of the file with the two ISO codes in your file's order, e.g. en\tpl — the importer will use that instead of guessing. (You can also skip the file and paste the rows straight into the box; it tells you how many cards it will import before you commit.)
That is the whole thing. Below is what to do if your deck is not a straightforward one.
If you changed the export checkboxes, or your deck is unusual
The five checkboxes. All off by default, and all of them should stay off. Include HTML and media references wraps every card in markup. Include tags, Include deck name, Include note type name and Include unique identifier each add a column — and the importer wants exactly two columns, so any one of them ticked means the file is rejected.
Why deleting the # lines matters so much. Leave them in and the import fails on every row, not just those two. The importer looks at the first line to decide whether your file is tab- or comma-separated, and #separator:tab says the word "tab" without containing an actual tab character — so it guesses comma, and then nothing lines up.
Decks over 1,000 cards. One import is capped at 1,000. A larger deck has to be split into several files and imported in batches. Most serious Anki decks are larger than this, and it is the single biggest reason the migration is more work than it should be.
Note types with more than two fields. If your cards have Front / Back / Example / Audio, the export has four columns and will be rejected even with every checkbox off. Open it in a spreadsheet, delete the columns you do not want, save it back out.
One bad row stops everything. The importer refuses the whole file rather than the offending lines, so the submit button stays disabled until it is clean. Annoying — though at least you never end up with half a deck.
Want to see the shape it expects? Here is an example spreadsheet you can copy — two columns, one card per row. The general version of this, with the CSV format spelled out, is in how to import content from CSV.
What happens to the cards after they land
This is the part that makes it worth doing, and it is the one thing Anki cannot do without a stack of add-ons.
Every imported card gets a voice — neural text-to-speech on both sides, in any of 76 languages, generated automatically. Every card also gets a phonetic transcription (IPA), produced automatically, including for scripts you cannot read yet. You do not install anything, configure anything, or pick a voice per deck. It simply happens, to all of them, including the twelve thousand you just brought over.
And once your cards have audio, they stop being only a thing you sit down to review. The same sentences play back like a podcast made entirely of your own material, on the commute or at the gym — that half of the app is here. If you would rather see how new cards get made in the first place, that is the Creator; how the audio and the transcription are actually produced is written out too.
Where Anki still wins
Worth saying plainly, because you will check anyway. Anki is free and open-source, and it works offline. Taalhammer is a paid subscription with a free trial, and it has no offline mode — on a plane or in a tunnel with no signal, it does nothing. If either of those matters to how you actually study, Anki is the better tool for you and no amount of automatic audio changes that.
What we are offering is the other trade: you stop maintaining a machine made of add-ons, and every sentence you own gets a voice and a transcription without you doing anything.
Free trial, then a subscription. No offline mode, and your Anki scheduling history does not come with you.