Speed Reader AI — Overview

Quickstart

1. Open the app and paste text, upload a file, or extract from a URL.
2. Select Visual, Read‑to‑Me, or AI Features from the Mode selector.
3. Press Start (Visual) or Play (TTS) to begin. Adjust speed, chunking, and ORP color as needed.
4. Use Export to save progress or Export as MD to produce a Markdown study sheet.

Controls reference

• Speed slider (WPM): Sets the display rate for RSVP in Visual mode. Typical comfortable range: 250–450 WPM. Advanced readers: 600–800+ WPM. Increase gradually to maintain comprehension.
• Words per chunk: Number of words shown per flash (1–5). 1–3 improves fixation and acuity at high WPM; 4–5 increases contextual coherence at lower WPM.
• ORP color: Optimal Recognition Point highlight color. Increase contrast for better fixation at higher speeds.
• TTS speed slider: Playback rate for Read‑to‑Me (0.25×–4.0×). For dense math/philosophy, use 0.75–1.25×; for lighter prose, 1.25–2.0×.

Reading modes & cadence

RSVP minimizes saccades (eye jumps) by keeping the text in a stable locus and controlling temporal cadence. Comprehension depends not only on WPM but on cadence shaping—how pauses and bursts map to syntax.

• Temporal chunking: Short function words tolerate higher cadence; content words benefit from micro‑pauses for integration.
• Punctuation heuristics (guidance): Pause longer after sentence terminators (., !, ?), medium after semicolons/colons, brief after commas/parentheticals. While the app streams at uniform WPM, you can simulate cadence effects by lowering WPM for dense passages and using sentence navigation to replay.
• Chunk size vs. WPM: At high WPM, use smaller chunks (1–2) to reduce masking; at lower WPM, larger chunks (3–5) restore prosodic feel.
• Fixation stability: Maintain consistent ORP color/contrast to anchor micro‑saccades and reduce regressions.

Visual speed reading

The visual reader displays one to five words at a time at a chosen WPM. Adjust chunk size, set the ORP color, and navigate sentence‑by‑sentence. A progress bar and live stats track words read, total words, time elapsed, and current WPM.

• Controls: Start, Pause, Reset, Fullscreen, Teleprompt mode
• Sentence navigation: Repeat current sentence, advance to next
• Stats: Words Read, Total Words, Time Elapsed, Current WPM

Teleprompt mode

Teleprompt presents the text as a large, smooth‑scrolling prompt for talks, presentations, or performance reading. While RSVP optimizes for speed at a fixed gaze, Teleprompt optimizes for delivery—breath, emphasis, and audience contact.

Purpose

• Rehearsal: Practice scripts with consistent pacing and visual clarity.
• Live delivery: Present with minimal cognitive load so attention can shift to tone and audience.

Cadence and phrasing

• Cadence: Aim for 150–180 WPM conversationally; 120–140 WPM for technical/novice audiences. Use sentence boundaries and clause commas as micro‑pause anchors.
• Phrasing: Break at semantic units (noun phrases, verb phrases) to maintain coherence. Avoid mid‑unit breaks that fragment meaning.
• Breath: Plan inhale points every 10–15 seconds; longer lines demand earlier breaths. Balance respiratory cadence with rhetorical emphasis.
• Prosody: Vary pitch and intensity at clause ends to signal structure. Technical density benefits from descending contours; persuasion can leverage rising contours on claims.

Typography and visibility

• Use large high‑contrast fonts and generous line height for peripheral readability.
• Keep line length 45–70 characters to reduce return‑sweep errors.

Read‑to‑Me (TTS)

Generate high‑quality speech from your text using selectable voices and adjustable playback speed. The player supports play/pause/stop, sentence navigation, and track downloads.

• Enter an API key for your selected provider
• Choose a voice and speed (0.25× – 4.0×)
• Style prompt: guide tone, pacing, and pronunciation
• Export: Full Track and Audio Pack (zip)

AI features (state‑of‑the‑art study tooling)

Speed Reader AI integrates with multiple providers to synthesize, structure, and test knowledge.

Summarize Text

• Abstractive focus: Produces concise, sectioned summaries capturing gist, claims, and caveats.
• Structure: Uses headings, bullet hierarchies, and key takeaways for rapid review.

Custom Summary

• Accepts style instructions (e.g., “for executives”, “Feynman explanation”, “compare X vs Y”).
• Useful frameworks: SCQA, Claim‑Evidence‑Reasoning, Problem‑Approach‑Result‑Implications.

Quiz Generator

• Active recall: Generates questions to retrieve knowledge, improving retention.
• Cognitive coverage: Mix factual, conceptual, procedural, and conditional questions (Bloom’s taxonomy from Remember→Evaluate).
• Calibration: Start with open‑ended prompts; add distractor choices to assess discrimination.

Document inputs

• Paste text directly into the editor with line numbers
• Upload supported files: .txt, .pdf, .epub, .docx
• Extract from a URL via the URL form
• OCR camera for scanned pages and photos

OCR camera

Built on Tesseract.js, the OCR camera captures text from images with adjustable levels to improve recognition. Use it to ingest printed material or whiteboard notes quickly.

Key terms

Double‑click words to collect them into your Key Terms panel. Print the list or clear it as needed. For best retention, convert terms into Q/A pairs (define, compare, apply) and interleave across sessions.

Chapter selection

For long documents, select chapters to process. Choose all or only the sections you want, then process the selected subset to focus your reading.

Export / Import

• Export progress to JSON; import later to resume
• Export as Markdown with headings, summaries, and notes
• Audio Pack export/import for TTS assets

AI providers & models

Speed Reader AI supports multiple providers. The app populates model lists dynamically from configuration. Below is a professionally curated overview of the options present in your configuration, including strengths, ideal use cases, and a pricing snapshot.

OpenAI

Frontier multimodal models with strong instruction‑following, tool use, and reasoning. Choose smaller variants for lower latency and cost.

Anthropic (Claude)

Strong at harmlessness and long‑form analysis with high factuality. Good for structured study materials and safer content generation.

DeepSeek

Competitive performance with a reasoner variant for chain‑of‑thought and tool‑use heavy tasks.

OpenRouter (aggregator)

Access many providers/models through a single API. Your config includes the following free or general routes:

Ollama (local)

Run local models with zero per‑token API cost and maximum data privacy. Performance depends on your CPU/GPU and chosen model size.

• Cost: $0 per API call (local compute only).
• Best when offline, cost‑sensitive, or requiring data residency.

// In-app steps:
// 1) Open AI Configuration
// 2) Select Provider (e.g., OpenAI)
// 3) Choose Model from the populated list
// 4) Paste API Key (if required)
// 5) Use Summarize / Custom Summary / Quiz in AI Features

Voices & speed

The TTS engine offers multiple voice options and playback speed control. Style prompt lets you specify tone, pacing, and pronunciation hints (e.g., acronyms, names).

Markdown exporter

Export your study materials in Markdown. The exporter structures headings and preserves lists for easy review in any Markdown viewer or notes app.

Readwise

Readwise integration helps consolidate your highlights and notes. Use the Readwise panel to sync summaries or key insights from your reading sessions.

Performance & accuracy

• Use chunking 1–3 words for maximum recognition at high WPM
• Increase ORP contrast to maintain fixation
• For OCR, ensure sharp images, good lighting, and straight alignment
• Alternate sprint and recovery blocks to maintain comprehension at higher speeds

Shortcuts

• ESC: exit fullscreen
• Space: pause/resume (visual)
• Arrow keys (where applicable): sentence navigation

Troubleshooting

• If TTS won’t start, verify provider, model, and a valid API key
• For PDFs that fail to parse, try OCR camera or export to .txt
• If sidebar highlighting desyncs, refresh once to reset anchors

Privacy

API keys are entered locally in the app. Review your provider’s data retention policy. Exported files remain on your device unless you choose to share them.

Changelog (high‑level)

• Visual reader with stats, sentence nav, fullscreen
• Read‑to‑Me with voices, speed, and style prompts
• AI features: summarize, custom summaries, quizzes
• OCR camera with adjustable levels
• Key terms, teleprompt mode, chapter selection
• Markdown and audio pack export/import