Why listen instead of read?
Reading a 40-page paper requires sustained visual attention. Your eyes tire, your focus drifts, and you end up re-reading the same paragraph three times. Listening is different. You can absorb the same content during a commute, a walk, or while doing dishes, without opening a single tab.
Studies of comprehension find that listening and reading draw on largely the same language skills. For explanatory text such as papers, textbooks and reports, many people understand about as much from listening as from reading, and hearing every word makes it harder to skim past the difficult parts.
Recitare's TTS is built on this premise: your reading list should not shrink just because your eyes are busy.
How does Recitare read documents aloud?
When you press play, Recitare does not send your raw file to a TTS API. It first parses the document into clean, ordered prose, removing page numbers, headers, footers, footnotes, and figure captions that would sound nonsensical when read aloud. Then it splits the text into individual sentences and feeds them to the TTS engine one by one.
Document parsing
PDFs are parsed in-browser using PDF.js with column detection and reading-order reconstruction. DOCX, TXT, PPTX, and images all go through the same clean-text pipeline. Web URLs are extracted and cleaned via the Readability algorithm.
AI text revalidation
For non-PDF documents, Gemini reorders any text fragments into correct human reading sequence before playback begins. This fixes cross-column merges and slide layout artifacts so nothing sounds garbled.
Sentence splitting
The cleaned text is split into individual sentences. Every sentence becomes one unit of highlighting, navigation, and TTS synthesis, so skip forward/backward works at the sentence level, not the paragraph or chapter level.
TTS synthesis
Each sentence is sent to the TTS engine appropriate for your plan tier. The engine returns an audio file (WAV or MP3) which plays through an HTMLAudioElement in the browser.
Sentence highlighting
As audio plays, the current sentence lights up in amber. On the Scholar tier, highlighting uses SSML timepoints for millisecond-accurate sync. On the Researcher tier, one sentence plays while the next is pre-loaded in the background for gapless delivery.
What voices are available?
Recitare offers six voice personas across three accents, with a male and female option for each. All six are available on the Scholar and Researcher tiers. Voice quality differs by tier. The underlying engine determines how natural and expressive the voice sounds.
- American Male
- American Female
- British Male
- British Female
- Australian Male
- Australian Female
On the Researcher tier, each voice persona is mapped to a native Inworld voice model with a persona-specific style instruction. This is what gives Researcher voices their warmth and expressiveness compared to generic TTS systems.
What is sentence-by-sentence highlighting?
As Recitare reads your document, each sentence lights up in amber the moment it begins playing, and the reading area auto-scrolls to keep it visible. When the next sentence starts, the previous one dims. This is the same interaction pattern as a karaoke display. Your eye knows exactly where in the document you are at every moment.
Scholar tier: SSML timepoints
Up to 8 sentences are batched into one API call wrapped in SSML markup. Google Cloud returns the MP3 plus a list of timepoints (e.g., sentence 3 starts at 4.217 seconds). The frontend polls the audio clock 60 times per second via requestAnimationFrame and advances the highlight exactly when each timepoint is crossed.
Researcher tier: pre-buffer pipeline
Each sentence is synthesized individually for maximum expressiveness. While one sentence plays, the next two are fetched and held in a buffer queue. When the current sentence ends, the next audio begins within milliseconds, no perceptible gap between sentences.
How is this different from a screen reader?
Screen readers are accessibility infrastructure. They read everything on screen: navigation menus, button labels, ARIA attributes, form fields. Recitare's TTS is purpose-built for document content:
| Screen Reader | Recitare TTS | |
|---|---|---|
| What it reads | Everything on screen | Document prose only |
| Removes noise | No | Headers, footers, page numbers, footnotes |
| Column order | DOM order (often wrong) | Spatially reconstructed reading order |
| Navigation | Element-level | Sentence-level skip forward / back |
| Voice quality | OS system voice | Inworld |
What is the difference between Free, Scholar, and Researcher?
Three tiers, three TTS engines. The fundamentals, sentence highlighting, skip controls, speed, voice selection, are identical across all tiers. What changes is voice quality, word limits, and the underlying synthesis technology.
| Free | Scholar · $9/mo | Researcher · $18/mo | |
|---|---|---|---|
| Engine | Inworld | Inworld | Inworld |
| Voice quality | High quality, natural | High quality, natural | Highest quality, expressive |
| Word limit | 1,000 / week | 100,000 / month | 300,000 / month |
| Voices | 6 AI voices | 6 AI voices | 6 AI voices + expressive style |
| Highlight sync | Sentence-level | Karaoke (SSML timepoints) | Sentence + 2-ahead pre-buffer |
| Gapless audio | No | Batch (8 sentences) | Yes (pre-buffer) |
What controls are available during playback?
Playback controls sit at the bottom of the reading area. On a phone, open Tools for speed, voices and additional controls.
Play / Pause
Tap to start or pause at any moment. Resumes from the exact sentence where you stopped.
Skip Forward / Backward
Jump one sentence at a time in either direction. Playback restarts from the new position.
Speed control
0.75x, 1x, 1.25x, 1.5x, 1.75x, 2x. Change speed mid-session; takes effect on the next sentence.
Voice selector
Switch between the 6 voice personas at any time. The new voice takes effect on the next sentence.
Voice engine
Recitare selects the voice engine according to your plan and remaining allowance. Choose a voice persona in Study tools, or Tools on a phone.
Stop
Fully stops playback and clears the audio buffer. Useful before switching documents or closing the app.
What technology powers the Researcher tier?
The Researcher tier runs on Inworld, released April 2026. It is independently ranked as the highest-quality text-to-speech engine available to developers, ahead of ElevenLabs, OpenAI, and Google WaveNet.
What makes it different from other AI TTS systems is native support for expressive audio tags. The system can embed natural-sounding emotion into speech: a concept explained with [enthusiastic] energy, a caveat delivered in a [thoughtful] tone. This is what makes the voice sound like a tutor rather than a text-to-speech robot.
The engine natively supports 14+ languages via the same voice personas. Recitare will read a French paper in French, a German journal article in German, without any configuration needed.
Who is TTS playback for?
Students with heavy reading loads
Get through assigned readings during your commute or gym session instead of losing evenings to a screen.
Grad students and researchers
Listen to literature review papers while conducting experiments or coding. Stop reading papers, start absorbing them.
Auditory learners
If you retain information better when you hear it rather than see it, TTS is the most natural way to consume written material.
Fast readers who want to go faster
Run TTS at 1.5x or 2x while following the highlighted text. Many users find this faster than reading, with better focus.
People with reading difficulties
Dyslexia, visual fatigue, ADHD: TTS with sentence-by-sentence tracking makes dense text accessible without accessibility compromises.
Non-native language readers
Hearing the pronunciation while following the highlighted text is a good way to learn academic vocabulary in a second language.
How do you start listening?
Four steps, under 30 seconds:
- Upload a PDF, DOCX, TXT, or PPTX, or paste a URL
- Wait for the document to parse (PDFs process in the browser, usually under 5 seconds)
- Select a voice in Study tools at the top right, or Tools on a phone
- Press play
No configuration. No choosing a reader mode. Just upload and press play. The Free tier starts immediately. Upgrading to Scholar or Researcher unlocks higher-quality voices and larger monthly word limits from the same interface.
Frequently asked questions
Is TTS playback free?
Yes. Every plan uses Inworld voices and all six personas. Free gives you a one-time 20,000-word welcome grant when you sign up (enough for a full paper or two), then 1,000 words per week. Scholar ($9/month) gives you 100,000 words per month. Researcher ($18/month) gives you 300,000 words per month on Inworld's highest-quality model, with more expressive delivery.
What voices are available?
Six voices across three accents: American male and female, British male and female, and Australian male and female. All six voices are available on every tier, Free, Scholar, and Researcher, powered by Inworld. Researcher tier upgrades to Inworld with expressive delivery.
What is karaoke-style word-by-word highlighting?
As Recitare reads your document, each sentence lights up in amber as it is spoken, then dims when the next sentence begins. On the Scholar tier, this highlighting is synced to the exact millisecond using SSML timepoints embedded in the audio. On the Researcher tier, sentences are pre-loaded one at a time with a 2-sentence pre-buffer for gapless delivery.
Can I change the reading speed?
Yes. All tiers support speed control from 0.75x (slow and deliberate) up to 2x (fast scan). The speed selector sits in the bottom playback bar, or Tools on a phone. Changing speed mid-playback takes effect on the next sentence.
How does skip forward and backward work?
The skip buttons move one sentence at a time. Skip forward jumps to the next sentence and resumes playback immediately. Skip backward goes to the previous sentence. If you are more than a couple of seconds into the current sentence, it replays from the start of it first.
How is this different from a screen reader?
Screen readers are accessibility tools designed to read any on-screen text including menus, buttons, and UI labels. Recitare's TTS is document-focused: it parses your full document into clean prose, removes headers, footers, page numbers, and figure labels, then reads the actual content. The result sounds like an audiobook, not a robot narrating a web page.
Does it work with all document types?
Yes. TTS works with PDFs (parsed in-browser via PDF.js), DOCX files, TXT, PPTX, images via OCR, and web content pasted as a URL. All formats go through the same sentence-splitting pipeline so highlighting and skip controls work identically regardless of source.
What happens if I close the tab mid-listen?
Recitare saves your reading position as you listen. Reopen the document to resume from the saved sentence or PDF reading position, or choose to start again. The most recent progress must have been saved before you close the page.