Skip to main content

​
Overview

Uxie’s text-to-speech feature turns your PDFs into audio, allowing you to listen while following along with synchronized highlighting. Choose from multiple voice engines and adjust playback speed to match your preferences.

​
Key Features

Word Highlighting

Follow along with word-by-word highlighting as the document is read

Multiple Voices

Choose from browser voices, Kokoro AI, or Supertonic AI voices

Speed Control

Adjust reading speed from 0.5x to 2x

Follow Along Mode

Auto-scroll the page to keep the current word in view

​
Getting Started

​
Starting Text-to-Speech

1

Open TTS Controls

Click the audio icon in the bottom toolbar
2

Press Play

Click the Play button to start reading from the current page
3

Follow Along

Watch as words are highlighted in real-time
4

Navigate

Use controls to pause, skip, or adjust speed
TTS requires the browser’s SpeechSynthesis API. All modern browsers support this feature.

​
TTS Controls

​
Playback Buttons

  • Play (▶): Start or resume reading
  • Pause (⏸): Temporarily stop without losing position
  • State persists - resume exactly where you paused
  • Skip (⏭): Jump to the next sentence
  • Useful for navigating quickly through familiar content
  • Maintains reading flow and highlighting
  • Stop (🚫): End reading session
  • Resets to beginning of current page
  • Clears all highlighting

​
Reading Speed

Click the speed button to cycle through speeds:
  • 0.5x - Slow, careful listening
  • 0.75x - Relaxed pace
  • 1x - Normal reading speed (default)
  • 1.25x - Slightly faster
  • 1.5x - Fast reading
  • 1.75x - Very fast
  • 2x - Maximum speed
Start at 1x and gradually increase speed as you get comfortable. 1.25x to 1.5x is optimal for most users.

​
Follow Along Mode

Toggle the Eye icon to enable/disable:
  • Enabled (highlighted): Page auto-scrolls to keep the current word visible
  • Disabled: Page stays fixed - you manually scroll
Perfect for:
  • Listening while doing other tasks
  • Following along in bed or on the couch
  • Studying without constant scrolling
Implemented at /src/components/pdf-reader/toolbar/tts-controls.tsx.

​
Voice Options

​
Voice Engines

Uxie supports three TTS engines:
Native system voices
  • Built into your operating system
  • No additional setup required
  • Fast and reliable
  • Quality varies by OS
  • Free
Available voices depend on your system:
  • Windows: Microsoft David, Microsoft Zira, etc.
  • macOS: Alex, Samantha, Victoria, etc.
  • Linux: eSpeak voices
AI-powered natural voices
  • High-quality neural TTS
  • More natural-sounding than browser voices
  • Requires WebGPU or WASM support
  • Multiple voice personas available
  • May require initial model download
Supported voices defined at /src/lib/tts/providers/kokoro-provider.ts
Premium AI voices
  • Studio-quality voice synthesis
  • Extremely natural prosody
  • Best voice quality available
  • May require API access
Supported voices defined at /src/lib/tts/providers/supertonic-provider.ts

​
Choosing a Voice

Voice selection UI is currently in development. The default voice is determined by your browser and system settings.

​
Reading Modes

​
Continuous Reading

Reading from start to finish:
  1. Navigate to your starting page
  2. Click Play
  3. TTS reads the entire page, then advances
  4. Continues until you stop or reach the end

​
Selected Text Reading

Read just a specific passage:
1

Select Text

Highlight the text you want to hear
2

Click Read Icon

Click the audio icon in the selection popover
3

Listen

TTS reads only the selected text
Implemented at /src/components/pdf-reader/highlight-popover.tsx:72.

​
Resume Reading

Continue from last position feature is in development. Currently, TTS restarts from the beginning of the current page when you reload.

​
Word Highlighting

​
How It Works

  1. TTS extracts text from the PDF in blocks
  2. Text is split into sentences
  3. As each word is spoken, it’s highlighted on the page
  4. Highlighting follows the audio in real-time

​
Highlight Appearance

  • Active word: Highlighted in bright color
  • Smooth transitions: Highlighting moves fluidly between words
  • Sentence-aware: Pauses briefly at sentence boundaries

​
Reading Modes

Two highlighting modes are available:
Standard word-by-word highlighting. Used when reading selected text or specific passages.
Sentence-by-sentence highlighting. Used for continuous document reading.
Mode constants defined at /src/components/pdf-reader/constants.ts:1.

​
Technical Details

​
Implementation

The TTS system consists of: Base Provider (/src/lib/tts/base-audio-provider.ts):
  • Abstract class for all TTS engines
  • Handles audio playback
  • Manages state (playing, paused, stopped)
Browser Provider (/src/lib/tts/providers/browser-provider.ts):
  • Uses Web Speech API
  • SpeechSynthesis interface
  • System voice access
Kokoro Provider (/src/lib/tts/providers/kokoro-provider.ts):
  • Neural TTS model
  • WebGPU acceleration
  • WASM fallback
Supertonic Provider (/src/lib/tts/providers/supertonic-provider.ts):
  • Cloud-based AI TTS
  • Premium voice quality

​
Engine Detection

export function getEngineFromVoice(voiceId: TTSVoiceId): TTSEngineType {
  if (BROWSER_VOICES.some((v) => v.id === voiceId)) {
    return "browser";
  }
  if (KOKORO_VOICES.some((v) => v.id === voiceId)) {
    return "kokoro";
  }
  if (SUPERTONIC_VOICES.some((v) => v.id === voiceId)) {
    return "supertonic";
  }
  return "browser"; // default
}
From /src/lib/tts/index.ts:14.

​
WebGPU Detection

For Kokoro AI voices: 
export async function detectWebGPU(): Promise<boolean> {
  if (typeof navigator === "undefined" || !("gpu" in navigator)) {
    return false;
  }
  try {
    const adapter = await navigator.gpu.requestAdapter();
    return adapter !== null;
  } catch {
    return false;
  }
}
WebGPU provides hardware acceleration for AI voice synthesis.

​
Reading Status States

enum READING_STATUS {
  IDLE,     // Not reading
  READING,  // Currently reading
  PAUSED    // Temporarily paused
}
State management ensures proper control flow and UI updates.

​
Best Practices

Use headphones: Better audio quality and less distraction for those around you.
Enable Follow Along: Let the page auto-scroll so you can focus on listening and understanding.
Adjust speed gradually: Start at 1x, then increase by 0.25x increments until you find your optimal speed.
Combine with highlighting: Listen while the AI reads, then highlight important passages afterward.
Use for editing: Listen to your own notes to catch awkward phrasing or errors.

​
Accessibility

TTS makes Uxie more accessible:
  • Visual impairments: Listen to documents without reading
  • Dyslexia: Hear correct pronunciation and pacing
  • Learning disabilities: Multi-sensory learning (audio + visual)
  • ESL learners: Improve pronunciation and listening skills
  • Multitasking: Absorb content while doing other activities

​
Limitations

  • TTS only works with text-based PDFs
  • Scanned PDFs must be OCR’d first
  • Some special characters may be mispronounced
  • Mathematical formulas are read as text, not equations
  • Tables may not read in logical order

​
Browser Compatibility

BrowserBrowser VoicesKokoro AISupertonic AI
Chrome✓✓ (WebGPU)✓
Edge✓✓ (WebGPU)✓
Firefox✓✓ (WASM)✓
Safari✓✗✓
Safari does not support WebGPU yet, limiting Kokoro AI voice availability.

​
Troubleshooting

  • Check system volume and browser permissions
  • Ensure speaker/headphones are connected
  • Try refreshing the page
  • Check browser console for errors
  • Verify SpeechSynthesis API support: open DevTools and run window.speechSynthesis
  • PDF may be image-based (use OCR first)
  • Some PDF formats don’t support text extraction
  • Try a different PDF viewer or re-export the PDF
  • Check if text is selectable in the PDF
  • Browser voices vary by OS and can sound robotic
  • Try Kokoro or Supertonic voices for better quality
  • Update your operating system for newer voice engines
  • On Windows, install additional language packs
  • PDF text extraction may have issues
  • Complex layouts (multi-column, tables) can confuse extraction
  • Try adjusting reading speed (slower can help)
  • Report the issue if it persists across documents
  • Ensure the button is highlighted (active)
  • Try toggling it off and on again
  • Check if page scrolling is locked by another extension
  • Refresh the page and try again

​
Future Enhancements

Planned features:
  • Voice selection UI
  • Bookmark positions to resume later
  • Download audio files
  • Customize highlight colors
  • Reading statistics (time listened, pages read)
  • Playlist mode (queue multiple documents)

PDF Reading

Navigate and view your documents

Annotations

Highlight while listening

OCR

Make scanned PDFs readable

Build docs developers (and LLMs) love

Get started for freeTalk to us