Meanings, not definitions
Tap a word or a whole phrase. The meaning is written for that exact sentence, not pulled from a dictionary entry or someone else’s saved definition.
Reading and SRS, finally the same app
Tap a word in one story.Meet it again in the next, on the day you’d forget it.
Real FSRS lives inside the reader. One tap builds the card out of the sentence you found it in, so there is no field to fill and no permanent yellow left behind. It comes back only when your odds of recall hit 90 percent.
Tap the brightest word
はころに。
Try the language you're learning
No fields to fill
Word, meaning, and source sentence become a scheduled card the moment you save. You never type a field.
Your reading is the review
Due words light up in the next thing you read and come back fainter as they get stronger. The page is the progress bar.
Language Lit closes the loop. Tap, saved, scheduled, met again in the next thing you read. Nothing to maintain, because there’s nothing left between the reading and the memory.
04
Meet it again
Real FSRS schedules each word from your review history. When one comes due, it lights up in whatever you’re reading. Fainter at each stage of memory: new, learning, growing, strong.
What power users notice
Saved words stay invisible in your reading until they’re due, so a highlight always means exactly one thing: your memory needs this word today. Nothing here is a status you assign or a color that sits on the page reminding you of work you already did. That’s also why progress feels physical. The text gets visibly cleaner as your memory gets stronger.
You don’t track your progress. You watch it happen on the page.
Why I built this
I didn’t need another app. I needed my reading and my reviews to be the same thing.
I moved to Tokyo excited to finally learn Japanese properly. I kept the streak for months. Then I turned on a normal show and caught maybe one word in ten.
So I did what serious learners do: real content, plus a stack to hold it up. A reader here, a dictionary there, flashcards on the side. It worked, and I hated it. I was spending more time feeding the system than reading, and every tool had a piece of the shape the research points at. None of them had the loop.
So I built the loop, and I use it every day. It’s the tool I wanted when I arrived in Tokyo, and I think it might be yours too.
Rom
Founder, Language Lit
Due-only sessions
Practice shows the words FSRS says are due, and nothing else. When you’re done, it says so. It never invents reviews to stretch a session.
Nothing to game
Both answers earn the same, so there’s no reason to flatter the scheduler. Your review history stays honest, and so does your schedule.
Progress is memory state
Highlight strength and Garden stages come from FSRS stability and retrievability, not from time spent in the app.
No pretend precision
Difficulty labels appear only where the text can be graded confidently. Where it can’t, the label stays blank instead of guessing.
The research this runs on
Sources last verified 2026-07-10
None of the method is our invention. Each finding below links to its source, so you can check the numbers instead of taking our word for them.
- 47.3% vs 36.7%
Across 254 studies and 14,811 participants, spaced practice beat massed practice 47.3% to 36.7% on recall. That gap is why a saved word waits, then returns on the day your memory actually needs it, instead of drilling while it is still fresh.
Cepeda et al. 2006, Psychological Bulletin, 132(3)Kang 2016, Policy Insights from the Behavioral and Brain Sciences, 3(1)
- 99.6%
In the open srs-benchmark, 10,000 Anki users and roughly 350 million real reviews, FSRS predicted recall more accurately than SM-2 for 99.6% of users. The same FSRS runs here, on your real review history, not a simplified imitation of it.
Open SRS benchmark, GitHub, open-spaced-repetitionYe, Su & Cao 2022 (KDD), Proceedings of the 28th ACM SIGKDD Conference (KDD ’22)
- d = 0.46
A meta-analysis of 34 extensive-reading studies, 3,942 learners, found reliable proficiency gains from reading real text: d = 0.46, rising to 0.71 in pre-post designs. Reading is the input engine of the whole loop, which is why the reader is the product.
Nakanishi 2015, TESOL Quarterly, 49(1)Jeon & Day 2016, Reading in a Foreign Language, 28(2)Krashen 2004, Libraries Unlimited
- 98%
Comfortable unassisted reading takes about 98% word coverage, roughly 8,000 to 9,000 word families. Tap lookups exist to close exactly that gap: they carry you through real text years before your vocabulary could carry you there alone.
Hu & Nation 2000, Reading in a Foreign Language, 13(1)Nation 2006, The Canadian Modern Language Review, 63(1)
- Shadowing
Following a text while hearing it improves listening comprehension and phoneme perception, with the largest gains for lower-proficiency learners. It is why every lesson has sentence-by-sentence audio you can slow to 0.5x and replay one paragraph at a time.
Hamada 2016, Language Teaching Research, 20(1)Kadota 2019, Routledge
Where we lose people
- You want something that listens to you. Nothing here does. No pronunciation score, no conversation partner, no speaking feedback of any kind.
- You love tuning card templates, add-ons, and scheduler settings. There’s nothing to tune here: cards are built one way, and FSRS runs on default weights, 0.9 retention, and fuzz on so reviews never pile into one brutal day. That’s the entire configuration, and you can’t change it. What you get back is nothing that breaks on update day.
- You’re at absolute zero in the language. Input-based reading needs a base to stand on, so you’ll get far more out of this once you have your first few hundred words.
Still here? Then this was probably built for you.
What it replaces
You were already running all of this.
Not a feature list. This is the list of jobs your habit already requires, and who does each one tonight.
The job
Your setup tonight
Here
A reader that remembers a word across every text
Your setup tonight: A separate subscription, or an install you maintain yourself
Here: Free, in all 33 languages
The meaning of a word in the sentence actually in front of you
Your setup tonight: A dictionary tab, plus your own call about which sense applies
Here: Free and unlimited, on every tier
Cards built out of what you read
Your setup tonight: Ninety minutes for every two hours of input
Here: One tap. You never type a field
A scheduler worth trusting with three years of review history
Your setup tonight: Free software, plus the add-ons, syncs, and templates you keep fixing
Here: Real FSRS, review logs kept, nothing to configure
Audio slow enough to shadow
Your setup tonight: A separate text-to-speech tool, or nothing
Here: Sentence by sentence at 0.5x, free
Progress you’d actually believe
Your setup tonight: A streak counter
Here: Stability and retrievability, drawn on the text itself
Every row lands on the free tier. Paid buys one thing that isn’t on this list: pointing all of it at your own books, videos, and files.
What you’re risking
Nothing, and here’s the mechanism.
Choose monthly or annual billing at signup. Change or cancel a paid plan anytime from account settings.
Straight answers
Is it actually FSRS, or "FSRS-inspired"?
Actual FSRS, scheduled from your real review history. Two answer buttons, Again and Good, because the middle buttons mostly add noise. Reviews appear only when due.
I have a serious Anki setup. Why switch?
Maybe don’t switch. Keep the deck you’ve built. Language Lit replaces the pipeline in front of it: lookup, card, and schedule happen in one tap while you read. Those words run on their own FSRS schedule here. Your old deck can keep doing its job.
How is this different from LingQ?
The short version: meanings are written for the exact sentence, not picked from community definitions. The scheduling is real FSRS, not an add-on. Due words come back inside your reading, not in a separate pile. The long version, feature by feature:
Read the full LingQ comparisonWhat is actually free forever?
Unlimited reading and tap lookups, full FSRS practice and the Garden, sentence-by-sentence audio, the 40,000-book library and public content, and 200 saved words. No card required. Paid adds importing and heavier AI.
Is my language supported?
The reader, lookups, FSRS, audio, and interface work in 33 languages. Japanese and Chinese use real segmentation, not tokenizer hacks. Automatic difficulty labels currently cover English, Spanish, Japanese, Korean, and Chinese.
Does it help with speaking?
Nothing here listens to you, so there is no pronunciation score and no conversation partner. What you do get is good shadowing material: every lesson has sentence-by-sentence audio in the narrator voice you pick, you can drop it to 0.5x, and you can replay a single paragraph until your mouth catches up. Plenty of people build output that way and get correction from a tutor. We’d rather draw that line here than after you sign up.
Save one word today. Meet it again this week.
Your first story is guided: three words, then a thirty-second review. That’s the whole loop, done, in one sitting. When one of those words comes back to find you in something new, you’ll know whether this is your tool. The cards were never the point. The book you actually want to read is.
Free forever · No credit card · 33 languages