Questions, answered plainly.

lull is an audio-first language-immersion app. You put in earbuds and listen to short stories generated in your target language at your exact level. When you hear a word you don't know, you tap once - anywhere on the screen. lull marks that word, schedules it for spaced repetition, and adjusts the difficulty of what comes next. Over time the unknown words shrink and review happens inside the listening rather than on a separate deck of flashcards.

First a short calibration estimates your level from a self-rating plus a frequency-banded comprehension probe. Then lull generates a short story using words you already know, a few new ones, and any words due for review. Every script is checked by trusted code against your allowed vocabulary, so the content genuinely sits at your level - it is not left to the model to behave. You listen; you tap the words you don't know; lull attributes each tap to the intended word, reschedules it with a spaced-repetition engine (FSRS), and tunes the next story up or down based on how often you tapped.

lull currently supports Brazilian Portuguese and Indonesian. Language is a first-class, swappable dimension built on a registry, so more languages can be added. If you already speak Spanish and are learning Brazilian Portuguese, lull gives you a head start by pre-seeding the large shared and cognate vocabulary the two languages have in common, so you start higher and only drill the genuinely different words.

Yes - that is the whole point. lull is audio-first and screen-optional. The default listening screen is a single blank tap target, so you can learn with earbuds in and your eyes closed, on a walk or a commute. Spaced review can also be done hands-available as quick audio-or-card flashcards when you have a moment to look.

Most apps are eyes-on-screen, gamified, and aimed at beginners, or they hand you a deck of flashcards to manage. lull is built for the intermediate plateau: comprehensible input (content just above your level), delivered by ear, with spaced repetition automated so you never build or grade a deck by hand. The single tap is the only input while listening. Reviews surface automatically when words come due.

Comprehensibility is enforced by code, not by trusting the language model. lull keeps a closed set of words you know plus a few new ones, and a trusted validator checks every generated script against that set, repairing any out-of-set words before you ever hear it. The model proposes; the validator decides. This is why the content reliably sits just above your level instead of drifting too hard or too easy.

Yes. You can point lull at a YouTube video or paste in lyrics or a transcript, and it will build a short lesson around it: it figures out which words are new for you, pre-teaches them with comprehensible warm-up stories, and then plays the real content. You learn toward the things you actually care about.

lull is in early access. You can try the web app today. Pricing for the iPhone app is still being finalized; there is no hard paywall during early access.