Words on Repeat
Turn anything you read into vocabulary you remember
3 followers
Turn anything you read into vocabulary you remember
3 followers
Paste a URL, upload a PDF, or import YouTube subtitles — AI extracts vocabulary flashcards with translations, example sentences, and grammar notes in seconds. Study with FSRS spaced repetition (the same algorithm Anki adopted in v24.11), quiz games, and daily goals. Unlike Anki: no learning curve. Unlike Quizlet: no $36/year paywall. Unlike both: AI builds your decks from real content you're already consuming. 185+ free decks in 12 languages. No ads, works offline. Free forever.
how does the AI handle niche or technical vocabulary from sources like research papers or industry-specific YouTube channels, does it actually get the translations right or do you end up correcting a bunch of cards?
@hikmet1329967 Hi Hikmet, good question — this is exactly where the approach matters.
The extraction runs on a large language model (via AWS Bedrock), not a dictionary lookup, and that's the key difference for technical vocab: it translates each word using the whole sentence it appeared in, not in isolation.
So terms from research papers or industry YouTube usually come out right, because the model has the surrounding context to disambiguate - it'll read "stability" in an FSRS paper differently than "stability" in a news article, and it recognizes domain jargon it's seen before.
Where you'll still want to eyeball it: genuinely obscure coinages, brand-new terminology, or ambiguous words in lower-resource languages. That's what the review step is for - before anything saves, you see every card (word, translation, example, grammar note) in a preview and can edit or drop any of them. For everyday and intermediate content I barely touch it; for dense technical material I'll tweak a handful, and the review flow makes that quick. So it's not "get a pile of cards and find the mistakes later" - you catch them up front in a few seconds.
If you've got a specific paper or channel in mind, paste it in and see how it does on your domain - genuinely curious, and happy to hear if anything looks off.