Recitare combines synchronized word-level listening with document-grounded explanations, comprehension checks, and study preparation. You stay connected to the original material, from the first page to the moment you need to demonstrate what you know.
attention-is-all-you-need.pdf
PII redactionWe propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely.
Max: What does the model lose, and gain, by dropping recurrence? [Page 1]
The problem
A PDF in one place. Generic AI in another. Notes somewhere else. And no clear connection back to the source. Recitare brings the whole process into one guided flow.
And a summary standing in for the understanding you actually needed.
One document. One flow. Always connected to the source.
Where you are in your studies
Catch up on the reading you are behind on, or go deeper on the reading you already do well.
See howWork through long reading lists and difficult theory without losing the argument.
See howFollow assigned English material, and clarify the hard parts in the language you think in.
See howThe Recitare method
Five connected stages, all inside the document you uploaded. You never leave the original material to get help with it.
Hear difficult academic material in a natural voice, at a speed suited to the reading.
Start the reading you have been putting off.
Word-level highlighting keeps your eyes on the original text as it is spoken.
Never lose your place in a dense argument.
Pause on a difficult term, model, or paragraph and ask for an explanation without leaving the page.
Get unstuck in seconds, not at office hours.
Document-grounded answers, page citations, and comprehension checks confirm you actually got it.
Know the difference between recognising words and understanding them.
Turn the reading into study guides, review questions, and flashcards drawn from the source.
Prepare from the material your course actually assigned.
Then demonstrate what you know with confidence.
In seminar discussion, in the assignment, in the presentation, and in the exam.
recitare: reader
PreviewDrop your paper
Parsed: 15 pages, 6,204 words
The dominant sequence transduction models are based on complex recurrent or convolutional neural networks.
Each head attends to the sentence from a different angle, so the model captures several relationships at once. [Page 4]
Max:Q, K and V are three “views” of your sentence, attention scores how much each word should look at every other word.
Cited from your diagramDrop in any paper, parsed in seconds
Follow
Recitare highlights each word as it is spoken, so you follow the original text rather than replacing it with detached audio. This is the difference between hearing a paper and actually reading one.
Choose your voice. All 6 available on every plan.
attention-is-all-you-need.pdf
PII redactionWe propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely.
Max: What does the model lose, and gain, by dropping recurrence? [Page 1]
Clarify
Pause on a hard term, model, or paragraph and get an explanation grounded in the document, with a path back to the page it came from. Switch modes on the same document, no re-uploading.
In development. Simplify will rewrite a difficult paragraph in plain language while you read, keeping the original text and its citations in view. Today you can ask Max to explain any passage in context.
How it will work
Original
“Attentional gating modulates encoding fidelity within the phonological loop.”
Simplified
“Focusing your attention helps your brain hold sounds more clearly.”
Hands-free playback with karaoke sync for commutes and long sessions.
ExploreMax asks Socratic questions and cites every answer to the source page.
ExploreAuto-generated quizzes and flashcards drawn straight from your material.
ExploreCross-reference, highlight, and export notes with a client-side toolkit.
ExploreUnderstand, then prepare
Max checks whether you can explain what you just read, shows you which ideas still need attention, then turns what you understood into preparation you can revise from.
Fig. 2: no caption, no mercy.
Max:Each “head” reads the sentence from a different angle, then the results are combined, like three readers highlighting the same page for different reasons.
Cited from your diagram, Page 4See the argument before deciding where to read more closely. Generated from your source, not the open web.
Spaced-repetition cards created from the concepts you actually listened to.
A two-host podcast-style discussion of your paper for a relaxed first pass.
Trust, by design
Every answer is grounded in the document you uploaded, not the open web, and points back to the page it came from. Concrete mechanisms, not a legal page of promises.
Names, emails, and identifiers are stripped before anything is sent to an AI model.
Every response from Max cites the exact [Page X] it came from, no unsourced claims.
Your documents are yours. They are never used to train models, full stop.
Data encrypted at rest and in transit, with an activity log of every AI action — visible to you.
Pricing
Recitare is not another general AI subscription. It is a specialized reading environment for getting your actual reading done, alongside whatever AI you already use. Lock in early-adopter rates while Recitare is in early access: your price stays the same for as long as you stay subscribed.
Everything you need to try real guided listening.
The plan for most readers: a month of course reading, guided and studied.
For grad students and intensive readers working across a substantial monthly literature load.
Secure payment via Stripe. Cancel anytime from your account.
Why Recitare exists
Recitare began as a side project in the first year of a PhD, born of a reading list that never got shorter. It is built around one idea: listen like you study, slowly, with a tutor beside you and a citation for every claim.
Independently built, now in early access. No investors telling us to chase skimmers instead of serious readers. Read the story
Recitare is in early access, so we won't invent quotes or numbers. Real words from real early users will live here. Want to be one of them?
Join the first cohortFAQ
Yes. When you sign up you get 20,000 free words to start, enough to listen all the way through a full paper (or two) with the same lifelike voices and word-by-word karaoke as the paid plans, no credit card. It's a one-time welcome grant that waits until you use it; after that, Free continues at 1,000 words a week.
PDFs, Word docs, PowerPoint, and images (we OCR them), plus any web article. You can also pull papers straight from arXiv, PubMed, and Project Gutenberg with one click.
No. Text-to-speech reads a document at you and stops there. Recitare keeps you inside the reading: each word is highlighted as it is spoken, you can pause on a hard passage and ask about it without leaving the page, answers cite the page they came from, and Max checks whether you can explain what you just heard before helping you build study material from it.
Listening gets you through material that felt resistant, which matters. But it is the start of the process, not the whole of it. Recitare is built so that listening leads somewhere: clarify what confused you, confirm you can explain it, then prepare from the original source. A summary is not the same as comprehension, and we do not present it as one.
Max only answers from the document in front of you and cites the exact [Page X] it drew from, no open-web guessing or invented sources. Anything personal is stripped out before it ever reaches a model.
Never. Your documents are yours. Personal details are auto-redacted before any AI call, everything is encrypted in transit and at rest, and every access is logged.
See how we handle your data →Yes. Subscribe during early access and your rate stays locked for as long as your subscription remains active. Price rises only ever apply to new subscribers, never to a continuously active plan.
Recitare runs in any modern browser on desktop and mobile, and a Chrome extension can read any webpage aloud. There is no app to install.
Six lifelike Inworld voices across American, British, and Australian accents, all with word-by-word karaoke highlighting, on every plan.
Yes, one click from your account. Billing is handled securely by Stripe.
Listen through it. Follow every word. Clarify what is difficult. Understand what matters. Prepare to demonstrate what you know.
Works in any modern browser: desktop & mobile · Chrome extension included
