About Tanzi

Tanzi is a learning tool for reading Chinese, not a translator. It explains the words you don't know where you meet them, in the apps you already use, then helps you learn them until you don't need the help.

It's built for the Chinese learners actually read: shop and order pages, social feeds, app settings and messages, usually in the middle of doing something else.

You point at a word and Tanzi shows its pinyin and dictionary meaning. A memory model then decides how much help to show next time and when you should review the word.

There are three ways to read: hold ⌥ and point at a word, select a passage to read it with pinyin, or press ⌘⇧2 and draw a box around Chinese in an image, scanned PDF or video. Words are split by longest match against CC-CEDICT, a dictionary of 125,132 entries built into the app. You get dictionary meanings word by word; it doesn't translate sentences or guess meaning from context.

Every lookup keeps its sentence. A word you look up in two sessions more than 30 minutes apart joins your reviews, up to 10 new words a day by default. FSRS, the memory model also built into Anki, estimates how well you know each word and schedules its review. The help on screen follows the same estimate: a full card for a new word, then pinyin and meaning, then a single line. Everything runs on your Mac, offline.

Because a few words make most of the difference, because help while reading works but isn't enough on its own, and because spaced reviews are remembered better than crammed ones. Three findings from reading and memory research shape the design.

First, a small set of characters makes up most Chinese text. In Jun Da's frequency list of modern Chinese, the 1,000 most common characters account for 89.1% of the characters in the corpus, and the 2,500 most common for 98.5%. Knowing a character doesn't always tell you the word it's part of, which is why Tanzi looks up whole words. And the last few unknown words in a sentence matter a lot. Writing about English readers, Paul Nation summarised the threshold:

“It was calculated that 98% text coverage (1 unknown word in 50) would be needed for most learners to gain adequate comprehension.”

So Tanzi aims its help at the few words in each sentence you don't know yet, and gives less as they become words you do.

Second, help while reading works, but most words still slip away. A meta-analysis by Yanagisawa, Webb and Uchihara of 42 studies with 3,802 learners found that people reading with glosses (meanings beside the text) learned 45.3% of the target words, against 26.6% without. On later tests it was 33.4% against 19.8%. Even with help, two-thirds of the words weren't learned. That's why Tanzi keeps every word you look up and brings back the ones you keep meeting for review.

Third, when you review matters. A meta-analysis by Cepeda and colleagues of 839 assessments in 317 experiments found that spaced reviews beat massed ones, and that the best gap grows with how long you want to remember:

“The ISI producing maximal retention increased as retention interval increased.”

FSRS sets each word's next review from your own history with it. The Anki manual says FSRS “can help you remember more material in the same amount of time”.

There's no launch date yet, and we won't make one up. The Mac comes first, for Apple silicon with macOS 15 or later. Android, iPhone and Windows come later.

It will be a one-time purchase with no subscription, and the price isn't set. We're still testing which apps it works in, and we'll publish the results before you pay for anything.

Tanzi is made by Go Baonic.

1. Jun Da, Middle Tennessee State University. Modern Chinese character frequency list, 2004.
2. I. S. P. Nation. The Canadian Modern Language Review 63(1). How Large a Vocabulary Is Needed for Reading and Listening?, 2006.
3. A. Yanagisawa, S. Webb and T. Uchihara. Studies in Second Language Acquisition 42(2). How Do Different Forms of Glossing Contribute to L2 Vocabulary Learning From Reading? A Meta-Regression Analysis, 2020.
4. N. J. Cepeda, H. Pashler, E. Vul, J. T. Wixted and D. Rohrer. Psychological Bulletin 132(3). Distributed Practice in Verbal Recall Tasks: A Review and Quantitative Synthesis, 2006.
5. Anki. Deck Options: FSRS.
6. MDBG and contributors. CC-CEDICT, 2026-09-28.

CC-CEDICT is licensed CC BY-SA 4.0. The bundled copy is the 28 September 2026 snapshot.