📚 Help Center

🚀 Getting Started

Type Shifter transforms plain text into beautifully formatted documents with optional Bionic Reading technology to help you read faster. Here's how to get started in minutes.

Quick Start Guide

👁️ Bionic Reading

Bionic Reading is a revolutionary reading method that can help you read up to 25% faster while maintaining or improving comprehension.

How It Works

Bionic Reading highlights the initial letters of each word (typically the first 2-4 letters) in bold. These bolded portions are called "artificial fixation points."

When you read text with Bionic Reading:

1. Your eyes are guided by the bold portions of each word
2. Your brain recognises these anchor points
3. Your mind automatically fills in the rest of each word
4. This reduces cognitive load and speeds up reading

Example

Standard text: "The quick brown fox jumps over the lazy dog."

Bionic Reading: "The quick brown fox jumps over the lazy dog."

Adjusting Intensity

The intensity slider (5% to 45%) controls what percentage of each word is highlighted:

• Low intensity (5-15%): Subtle highlighting, just the first 1-2 letters. Good for experienced readers who want minimal visual guidance.
• Medium intensity (20-30%): Recommended starting point. Provides clear guidance without being overwhelming. Most users find their sweet spot here.
• High intensity (35-45%): More letters highlighted per word. Helpful for very long words, unfamiliar content, or readers who need stronger visual anchors.

Who Benefits Most

• People with ADHD: The fixation points help maintain focus and reduce mind-wandering
• Readers with dyslexia: Combined with OpenDyslexic font, provides significant reading improvement
• Students: Process textbooks and study materials faster
• Professionals: Get through reports, emails, and documents more efficiently
• Non-native speakers: Visual anchors help with unfamiliar vocabulary
• Anyone who reads extensively: Reduce eye strain and reading fatigue

🎨 Style Templates

Type Shifter includes 60 professionally designed templates, each optimised for different types of content and reading contexts.

Compare Mode

Enable Compare Mode to view your original text alongside the transformed version. This is useful for:

• Seeing exactly what changes Type Shifter made
• Comparing readability before and after
• Quality checking before export
• Demonstrating Bionic Reading to others

🔊 Listen & Save Audio

Type Shifter reads your documents aloud with 28 high-quality neural voices and lets you save the audio as MP3 files. The whole feature runs inside your browser using a small neural model called Kokoro-82M, so your text never leaves your device and there is no monthly fee, no API key, and no usage limit. Listen to a research paper while you commute, save a long article as an audiobook for the gym, or have a draft of your own writing read back to you for proofreading.

What Powers the Listen Feature

Behind the scenes, Type Shifter uses three components that all run locally:

• Kokoro-82M: An open-source neural text-to-speech model. 82 million parameters, distilled from larger models for fast on-device synthesis. Quality is roughly comparable to commercial offerings like Amazon Polly Neural and Microsoft Azure Neural.
• WebGPU / WebAssembly runtime: Runs the neural model in your browser. WebGPU uses your graphics card for acceleration; WebAssembly uses your CPU as a fallback. The system picks the faster option automatically.
• lamejs MP3 encoder: A pure-JavaScript MP3 encoder. Compresses Kokoro's raw audio into MP3 files that play anywhere.

The Listen Panel

Look directly above the document preview canvas. You'll find a cyan-bordered panel with several controls. The controls reveal themselves as you use the feature: pick a voice and click Listen, and the save buttons appear.

• Voice dropdown: 28 neural voices grouped by accent and gender.
• Speed slider: Adjust playback from 0.5x (slow) to 2.0x (fast). Default is 1.0x.
• 🔊 Listen (cyan): Start reading the formatted text aloud. Also pauses and resumes mid-playback.
• ⏹ Stop (red): Stop playback immediately and cancel any audio still being generated.
• 💾 Save Recording (green): Save what's been played so far as an MP3.
• 📥 Save Full Doc (purple): Generate the entire document as an MP3 in the background.
• 🖍 Save Selection (orange): Save only the highlighted portion as an MP3.

The 28 Voices in Full

All 28 voices are English, but spread across two accents and both genders so you can pick something that matches the document's tone or your personal preference. Switching between voices is instant because they all share the same model file.

🇬🇧 British Female (8 voices)

• Emma: Calm, conversational, warm. The default British voice. Excellent for general reading and audiobooks.
• Isabella: Slightly brighter and more youthful than Emma. Good for blog posts and articles.
• Alice: Soft, gentle, slightly slower delivery. Excellent for poetry and reflective writing.
• Lily: Bright and cheerful. Best for short, upbeat content like marketing copy or social posts.
• Charlotte: Formal and precise. Reads journals, legal text, and academic papers clearly.
• Sophia: Mature, authoritative tone. Excellent for business writing and news.
• Olivia: Neutral, professional. A safe choice for technical documents.
• Mia: Energetic, conversational. Best for storytelling and informal writing.

🇬🇧 British Male (6 voices)

• Daniel: Calm, measured, slightly deep. The default British male voice. Excellent for non-fiction.
• Henry: Authoritative and clear. Best for news, research papers, and documentaries.
• James: Warmer and friendlier than Henry. Best for fiction and personal essays.
• Oliver: Slightly faster pace, brighter tone. Best for tutorials and how-to content.
• William: Mature, gravelly. Best for historical content and serious analysis.
• George: Mid-range, neutral. A reliable all-purpose male voice.

🇺🇸 American Female (8 voices)

• Bella: Particularly natural-sounding, expressive intonation. A standout voice for fiction and casual content.
• Nicole: Bright, clear, energetic. Best for marketing and educational content.
• Sarah: Warm, friendly, slightly conversational. Best for blog posts and personal writing.
• Sky: Soft and gentle. Best for meditation, poetry, and slow reads.
• Adam: Versatile, neutral tone (despite the name, listed as female in the Kokoro voice set). General-purpose.
• Heart: Emotive, expressive. Best for narrative fiction and emotional writing.
• Bonbon: Cheerful, upbeat. Best for short, light content.
• Aurora: Calm and slightly ethereal. Best for reflective and philosophical writing.

🇺🇸 American Male (6 voices)

• Michael: The default American male voice. Clear, neutral, professional. Excellent for almost any document type.
• Eric: Casual, conversational. Best for blog posts and personal essays.
• Liam: Bright, slightly youthful. Best for tutorials and how-to content.
• Onyx: Deeper, more resonant. Best for serious non-fiction, documentaries, and audiobooks.
• Echo: Mid-range, neutral. A reliable all-purpose American male voice.
• Fenrir: Authoritative and gravelly. Best for historical and dramatic content.

The Speed Slider

The speed slider sits next to the voice dropdown and adjusts playback rate from 0.5x (half speed) to 2.0x (double speed). The default is 1.0x (natural pace). Speed changes apply mid-playback without restarting, so you can drag the slider while you're listening and the rate adjusts smoothly.

Practical rates:

• 0.5x to 0.75x: Useful for language learning, poetry, or technical content where you want to absorb each word.
• 1.0x: Natural speaking pace. Good for fiction and content you're hearing for the first time.
• 1.25x to 1.5x: The audiobook sweet spot. Most listeners find this comfortable for non-fiction. Saves significant time on long content.
• 1.75x to 2.0x: Fast skim mode. The voice stays intelligible but starts to sound rushed. Best for material you already know.

The speed setting is applied at playback time, not generation time. So when you Save Full Doc, the speed control is irrelevant for the saved file; the MP3 is always generated at natural pace and your playback app (Apple Books, Pocket Casts, etc.) can speed-shift the file itself.

WebGPU vs WebAssembly

Generating neural-quality voice in real time is computationally expensive. Type Shifter uses whichever acceleration path your browser supports best, automatically.

WebGPU (faster, modern browsers with a graphics card)

WebGPU is the new browser standard for running computations on the graphics card. When available, Type Shifter uses it to generate audio roughly 5 to 10 times faster than the CPU path. A one-hour audiobook generates in about 5 to 8 minutes.

WebGPU is supported in:

• Chrome 113+ on Windows, macOS, ChromeOS
• Edge 113+ on Windows, macOS
• Firefox 121+ on Windows (experimental on macOS and Linux)
• Safari 18+ on macOS 15 and iOS 18 (newer Apple devices)

WebAssembly (universal fallback)

If your browser doesn't support WebGPU, or your GPU isn't compatible, Type Shifter falls back to WebAssembly. This runs the same neural model on the CPU instead. Quality is identical; the only difference is speed. A one-hour audiobook generates in about 30 to 50 minutes.

You can check which mode is active by opening the browser's developer console (F12 in most browsers, then click the Console tab) while pressing Listen. If you see "WebGPU initialised", you're on the fast path. If you see "WebGPU failed, falling back to WebAssembly", your hardware doesn't support GPU acceleration, but the feature still works.

Step-by-Step: Listen to Your Document

Pause and Resume vs Stop

The Listen button doubles as pause and resume. The Stop button is separate and behaves differently. Knowing the difference saves frustration on long generations.

• Pause (Listen button while playing): Click 🔊 Listen during playback to pause. The current chunk finishes playing, then audio stops. The next chunk stays loaded in memory.
• Resume (Listen button while paused): Click 🔊 Listen again to continue from where you paused. No re-generation needed.
• Stop (Stop button): Click ⏹ Stop to cancel playback entirely. Any chunks still being generated are discarded. Pressing Listen again starts over from the current cursor position.

Rule of thumb: pause for short breaks (answering the door, taking a sip of tea), stop when you're done or want to start over with different settings.

Saving Audio as MP3

Three save buttons sit alongside Listen and Stop. They serve different use cases. They appear once there is text in the preview and a voice selected.

💾 Save Recording (green)

Saves whatever has been read aloud during the current Listen session. The save button captures exactly what you heard, including any partial chunks at the end if you stopped mid-sentence.

• When to use it: You started listening, decided you wanted a saved copy of what you'd heard so far, and clicked save before finishing.
• What you get: An MP3 of every chunk that was played, in order. If playback finished, you get the whole document.
• When it's enabled: As soon as at least one chunk has finished playing. Until then, the button stays greyed out.
• Filename: The MP3 is named after the first few words of the document, with the date appended. For example: type-shifter-the-quick-brown-fox-2026-05-15.mp3.

📥 Save Full Doc (purple)

The most powerful save option. Generates the entire document silently in the background without playing audio, then downloads the resulting MP3 automatically when finished. You don't need to sit through the playback or keep the audio playing.

• When to use it: You want a complete audiobook of a long document to listen to later, offline, on a different device. Or you want all of it but don't want to sit through 47 minutes of TTS.
• Live progress counter: The button text updates to show progress: 📥 7 of 47, then 📥 8 of 47, and so on. Once two or three chunks have been generated, an estimated time remaining appears above the button: "About 4 minutes left".
• What happens during generation: Your browser uses your GPU (or CPU on WebAssembly). You can still use the rest of the app for things that don't need the GPU. You can change tabs, but keep the Type Shifter tab open so the generation doesn't get throttled. If you close the tab, generation aborts.
• How long it takes: Roughly 1 second of audio per 1 second of WebGPU generation (1:1 ratio), or 1 minute of audio per 1 minute of WebAssembly. So a 47-minute audiobook needs ~47 minutes on WebGPU, ~3-4 hours on WebAssembly. Plan accordingly.
• What you get: A single MP3 file containing the entire document, gapless between chunks.
• Mobile note: Save Full Doc works on phone browsers but is much slower because most phones don't expose WebGPU. For long documents, generate on a laptop and transfer the file.

🖍 Save Selection (orange)

Saves only the text you've highlighted in the preview canvas. Useful for pulling key passages out of a long document without saving the whole thing.

• When to use it: You want just the abstract of a research paper, the conclusions section of a report, a specific quote, or a particular chapter of a longer document.
• How to highlight: Click and drag across the text in the preview canvas (same as selecting text anywhere on the web). Touch and drag on mobile. The selection can span multiple paragraphs.
• What happens with partial sentences: If your selection cuts off mid-sentence, the audio reads exactly the selected portion. The voice doesn't try to be clever and round to the nearest sentence; what you highlight is what you get.
• Why this exists: Faster than Save Full Doc for short clips, and gives you finer control than waiting for Listen to reach the bit you want.

Step-by-Step: Make an Audiobook of a Long Document

Audio Quality and File Specs

128 kbps is high quality for spoken word. Music typically wants 192-320 kbps for transparency, but speech only uses a narrow frequency range and 128 kbps captures it cleanly without bloating file size.

Where MP3 Files Are Saved

The MP3 saves to your browser's default download folder, exactly like any other file you'd download from a website. Specifically:

• Windows: C:\Users\YourName\Downloads\
• macOS: ~/Downloads/
• iOS Safari: Files app → "Downloads" folder (or wherever your default download location is set)
• Android Chrome: Internal Storage → Download folder, also visible in the Files app

Your browser may prompt to confirm the download. If you've disabled the "Ask where to save each file" setting, the MP3 saves silently to your default Downloads folder.

Combining Listen with Other Features

The most useful Type Shifter workflows combine Listen with other features:

• Listen + Bionic Reading: Listen to a document while also seeing the Bionic-formatted version on screen. Both the voice and the visual cue reinforce each other, useful for active comprehension.
• Listen + Dark Mode: Save your eyes at night while still consuming content. Great for evening reading sessions.
• Listen + Compare Mode: Hear the original text while comparing it visually against your edited version (or vice versa).
• Listen + OCR: Photograph a printed book or article, OCR it, then save as an audiobook. Effectively turns any printed material into a personal audiobook.
• Listen + Speed: Use 1.5x for non-fiction to save time, drop to 0.75x when you encounter technical sections you want to absorb carefully.

Limitations and Known Issues

• English only currently. All 28 voices are English, with British or American accents. Other languages may come in a future update but are not supported today.
• No voice cloning. You cannot upload a sample of your own voice to clone it. The 28 voices are the only options.
• Numbers and abbreviations. Most numbers, dates, and common abbreviations are pronounced correctly, but obscure abbreviations (industry-specific jargon, drug names, place names) may be read letter-by-letter. Edit the input to spell them out phonetically if needed.
• Punctuation matters. Kokoro uses punctuation to determine pacing and intonation. Documents with poor punctuation (missing full stops, run-on sentences) will sound rushed or unnatural. Add punctuation before generating if needed.
• Very long sentences. Sentences over ~250 characters get split into multiple chunks at the nearest punctuation. This is invisible in playback but means very long sentences without commas may sound slightly choppy at the split points.
• Mobile WebGPU. Most phone browsers don't yet expose WebGPU, so generation is slow on phones. Save Full Doc on a laptop and transfer the file for the best experience.

📷 Image OCR (Scan Photos and Screenshots)

Type Shifter can extract editable text from any image you drop into the upload zone. Photos of book pages, screenshots of articles, scanned PDFs, printed letters, recipes from a cookbook, even photographs of street signs. The recognition runs entirely inside your browser using Tesseract.js with the high-accuracy English data file, so your images never leave your device, there's no usage limit, and it works offline after the first run.

OCR (Optical Character Recognition) is the technology that converts pictures of text into actual machine-readable text. Done well, it bridges the gap between paper and digital, turning a 200-page printed book into something you can search, edit, listen to, or convert into other formats.

What Powers the OCR Feature

Three components work together to make in-browser OCR fast and accurate:

• Tesseract.js: A JavaScript port of Google's open-source Tesseract OCR engine. Version 5 with LSTM neural network recognition, the same engine used in Google Drive's OCR scanning. Apache 2.0 licensed.
• tessdata_best (English): The highest-accuracy English language data file available. About 50 MB compared to ~10 MB for the default file. Roughly five times larger, but significantly more accurate on difficult images.
• Image preprocessor: A Type Shifter pre-processing step that prepares each image before sending it to Tesseract. 2x upscale, greyscale conversion, and contrast boost to make text edges sharper.

How to Use OCR

Which Images Work Best

OCR quality depends heavily on the source image. From the recogniser's point of view, what matters is contrast (dark text on light background, or vice versa), resolution (text at least 30 pixels tall), sharpness (no motion blur), and geometry (text aligned, not at an angle).

Here's the rough order from best to worst:

• Excellent (95%+ accuracy): Screenshots of articles or webpages, scanned PDFs of typeset documents at 300+ DPI, photos of book pages taken straight-on with good lighting on a clean white background.
• Good (85-95% accuracy): Printed letters, magazine articles, journal papers, photos of receipts with clear printing, well-lit signs.
• Variable (60-85% accuracy): Newspaper photos (multi-column layouts can confuse the recogniser), photos taken at a slight angle, lower-resolution scans, slightly faded prints, photos with mixed lighting.
• Poor (under 60% accuracy): Cursive handwriting, faded or thermal-printed receipts (the kind that fade after weeks), very small text (under 30 pixels tall in the image), text overlaid on busy graphics, severely angled photos, motion-blurred images.
• Unsupported: Pure handwriting (Tesseract is trained on printed type), languages other than English (English-only currently), mathematical equations and scientific notation, and image-only logos where the "text" is really a stylised graphic.

How to Photograph Books and Pages Well

If you're photographing printed material, a few minutes of technique training dramatically improves OCR accuracy:

• Lighting: Use natural daylight or two soft lamps to eliminate shadows. A single light source casts a hand shadow that confuses the recogniser. Avoid yellow indoor light if you can; daylight or daylight-balanced LED produces cleaner results.
• Angle: Hold the phone directly above the page, parallel to the page surface. If you're at an angle, the text on the far side of the page becomes smaller in the image and OCR accuracy on that side drops sharply.
• Distance: Fill the frame with the page (not your whole desk and a corner of the page). Closer means more pixels per character, which means more for Tesseract to work with.
• Focus: Tap on the text to set focus before shooting. Most phone cameras default to centre-focus, which is usually fine, but for safety tap to confirm.
• Stability: Use a tripod, lean your elbows on the desk, or rest the phone against a stack of books. Motion blur is the most common quality killer.
• Page flatness: Press down the book pages flat if you can, or photograph one page at a time rather than a spread. Curved page edges produce distorted text near the spine.

What Type Shifter Cleans Up Automatically

Raw OCR output is rarely clean. Every visual line in the image becomes a separate paragraph, hyphenated words get split across lines, and screenshot edges often contain garbled icon characters. Type Shifter applies several cleanup passes automatically before the text reaches the input area:

• Chrome-line trimming: Garbled status bar text and navigation icons at the very start and very end of the recognised text are detected and removed. The detector looks for lines that are mostly symbols (battery icons, time stamps, wifi bars) rather than letters. The body of the document is left untouched.
• Hyphenation rejoin: When a word is split across lines with a hyphen (like "pri-" on one line and "mary" on the next), the hyphen is removed and the word rejoined. This handles the typesetter's hyphenation convention found in books and printed articles.
• Soft-wrap collapse: Single line breaks within a paragraph (the visual line wraps from the image) become spaces, so sentences flow naturally instead of fragmenting. Without this step, a single sentence might span ten lines in the textarea.
• Paragraph preservation: Double line breaks (genuine paragraph boundaries in the source image) are kept intact. So you still get distinct paragraphs in the output, just with each paragraph flowing as one piece.

All four steps run automatically. There's no setting to disable them, but they're conservative. They only act on patterns that are clearly typesetting artefacts rather than intentional formatting.

Common OCR Errors to Look For

Tesseract is excellent but not perfect, especially on lower-quality source images. After scanning, skim the input area for these common mistakes:

• Number and letter swaps: "0" for "O", "1" for "l" or "I", "5" for "S", "8" for "B", "6" for "G". Most common in monospace fonts (computer printouts) or sans-serif fonts where letter shapes are simpler.
• Missing punctuation: Full stops, commas, and apostrophes sometimes drop out on lower-resolution images. The recogniser is conservative about punctuation it isn't sure of.
• Joined or split words: "twowords" instead of "two words" (happens when the space is narrow), or "p arsley" instead of "parsley" (happens when a character has a small internal gap).
• Capitalisation drift: Headings in decorative fonts sometimes come out with mixed case. "CHAPTER 1" might become "ChAPTER 1" or "ChApTer 1".
• Mis-recognised similar characters: "rn" looks like "m", "cl" looks like "d", "vv" looks like "w". Look out for these in the words.
• Apostrophes flipped: Curly typographer's quotes (' ') sometimes come out as straight quotes ('), and contractions like "don't" become "don t" if the apostrophe is unclear.
• Bullet points and list markers: Some bullet glyphs come through as letters or symbols rather than •. Easy to fix manually.

Fix these in the input area before clicking SHIFT MY TEXT or Listen. The textarea is fully editable, so you can use keyboard shortcuts like Ctrl+F (or Cmd+F on Mac) in some browsers to find and replace.

The Image Preprocessing Pipeline

Before sending your image to Tesseract, Type Shifter runs three preprocessing steps in a small canvas. These dramatically improve recognition accuracy on real-world photos and screenshots:

1. 2x upscale. The image is scaled up to twice its original dimensions using bilinear interpolation. Tesseract is happiest with characters that are at least 30 pixels tall; upscaling small text first gives the recogniser more pixels to work with.
2. Greyscale conversion. Colour is discarded. OCR doesn't use colour information, so removing it speeds up processing and prevents background colours from interfering with character recognition.
3. Contrast boost (1.4x). A contrast multiplier of 1.4 is applied, deepening blacks and brightening whites. This sharpens text edges and reduces grey "halo" around characters in low-contrast images.

These run automatically. There's no setting to adjust them, but they're tuned for the common case of "photo of printed page" or "screenshot of webpage" and rarely hurt.

Page Segmentation Mode

Tesseract has different modes for how to interpret the layout of an image. We use mode 6 (PSM 6), which assumes the image contains a single block of uniform text. This is the right choice for most book pages, articles, and screenshots.

If you're scanning a complex layout (multi-column newspaper, a heavily-formatted brochure, a webpage with sidebar and main content), the recogniser may try to read the columns in unexpected order. For best results on complex layouts, crop the image to a single column before OCR, then paste each column separately.

Combining OCR With Other Features

The most useful Type Shifter workflows combine OCR with other features:

• OCR + Listen + Save Full Doc: Photograph a book chapter, OCR it, save as MP3, listen on your commute. Effectively turns any printed material you own into a personal audiobook. Particularly useful for textbooks, academic papers, and out-of-print books that aren't available as audiobooks.
• OCR + Bionic Reading: Scan a difficult-to-read photograph or faded document, then turn on Bionic Reading on the result. Particularly helpful for dyslexic readers and ADHD users who struggle with dense printed material. The OCR makes the text editable, and Bionic Reading makes it visually easier to track.
• OCR + Templates: Scan a recipe from a cookbook, apply the Kitchen Recipe template, export as PDF. Suddenly your printed-only recipes become digital, searchable, shareable, and printable in your preferred format.
• OCR + Export to DOCX: Convert scanned PDFs (image-only) into proper editable DOCX files. Useful for older documents, archived PDFs that were saved as images rather than text, or printed letters you want to digitise for editing.
• OCR + Dark Mode: Read photographed material on screen at night without the original page's bright white background straining your eyes. The OCR'd text uses Type Shifter's dark theme regardless of the source image colours.
• OCR + Save Selection: Photograph an entire book page, OCR it, highlight just the paragraph you want, save as MP3. Perfect for pulling key quotes out of longer printed material.

Why English Only (For Now)

Type Shifter currently ships with only the English language data because each additional language adds another ~50 MB download. Bundling all of Tesseract's supported languages would balloon the first-use download to several hundred megabytes, which is impractical for a web app.

Tesseract itself supports over 100 languages including French, Spanish, German, Italian, Dutch, Portuguese, Russian, Chinese (Simplified and Traditional), Japanese, Korean, Arabic, Hindi, and many more. A future update may add an optional language picker that downloads additional language data on demand. For now, English-only.

First-Use Performance Notes

The first time you use OCR, your browser downloads two assets:

1. The Tesseract.js engine itself (~2 MB JavaScript bundle).
2. The high-accuracy English language data, also known as tessdata_best for English (~50 MB).

Both get cached permanently in your browser's IndexedDB storage. Subsequent OCR runs skip these downloads entirely, so all you wait for is image preprocessing (1-2 seconds) and actual recognition (5-15 seconds for an average page).

We deliberately chose the higher-accuracy data file (about five times larger than the default tessdata_fast) because the quality difference is significant on real-world images. The default file works fine for very clean scans, but stumbles on photos, faded prints, and unusual fonts where the better file shines.

OCR Limitations and Known Issues

• English only currently. See above.
• Handwriting is not supported. Tesseract is trained on printed type. Cursive and most handwriting will produce gibberish.
• Mathematical equations. Symbols, fractions, integrals, and Greek letters often come out wrong. OCR for math is a specialised field; consider tools like Mathpix for equation-heavy content.
• Multi-column layouts. Tesseract reads top-to-bottom, left-to-right. Multi-column newspapers may produce text where columns interleave unexpectedly. Crop to one column at a time for best results.
• Tables. Table structure is not preserved. Tables come out as plain text with the cells flowed together.
• Images within images. OCR ignores graphics and reads only the printed text, but illustrated content (textbooks, comics) may produce text that's missing context.
• Rotated text. Text rotated 90 degrees (like book spines) is not auto-detected. Rotate the image before uploading.
• Very small text. Text under ~30 pixels tall in the image is unreliable. Re-photograph closer if the result is poor.

📁 Import & Export

Supported Import Formats

Type Shifter can read the following file formats:

• TXT - Plain text files. The simplest format, works universally.
• RTF - Rich Text Format. Preserves basic formatting from word processors.
• DOCX - Microsoft Word documents (2007 and later). Extracts text content while preserving paragraph structure.
• PDF - Portable Document Format. Text is extracted from the PDF (note: scanned/image PDFs may not extract properly).
• MD - Markdown files. Popular with writers and developers. Markdown formatting is converted to proper headings and lists.
• EPUB - E-book format. Extract text from e-books for reformatting.
• HTML - Web pages. Useful for reformatting online articles.

Export Formats

Export your transformed documents in these formats:

• HTML - Best for web viewing
  Opens in any browser. Preserves all formatting including dark mode. Ideal for sharing online or viewing on any device. Smallest file size.
• DOCX - Best for editing
  Opens in Microsoft Word, Google Docs, LibreOffice. Allows further editing. Good for documents you need to modify or collaborate on.
• EPUB - Best for e-readers
  Works with Kobo, Apple Books, and most e-readers. Can be converted to Kindle format. Perfect for reading long documents on dedicated e-ink devices.
• PDF - Best for printing and sharing
  Universal format that looks the same everywhere. Ideal for printing, archiving, and sharing when you don't want the content edited.

♿ Accessibility Features

OpenDyslexic Font

OpenDyslexic is a free, open-source typeface specifically designed to increase readability for people with dyslexia. It's included in Type Shifter and can be enabled with one click.

Key features of OpenDyslexic:

• Weighted bottoms: Each letter has a heavier bottom, which helps prevent visual "flipping" or rotation - a common issue for dyslexic readers.
• Unique letter shapes: Similar letters like b, d, p, and q have distinct shapes to prevent confusion.
• Increased spacing: More space between letters reduces visual crowding.
• Consistent baseline: Helps eyes track along lines more smoothly.

Who benefits:

• People with dyslexia (affects approximately 10% of the population)
• Readers who experience visual stress or fatigue
• Anyone who finds standard fonts difficult to read
• Students and professionals who read for extended periods

Custom Fonts

Type Shifter includes over 1,900 Google Fonts you can apply to your documents, with separate controls for heading and body fonts:

• Sans-Serif: Including Inter, Roboto, Open Sans, Lato, Montserrat, and hundreds more
• Serif: Including Merriweather, Playfair Display, Lora, Georgia, and many more
• Display & Handwriting: Including Abril Fatface, Bebas Neue, Caveat, Cinzel, and many more
• Monospace: Including Fira Code, JetBrains Mono, Source Code Pro, and more

Custom fonts are found in the "Customize Sizes & Colours" section. Note: When OpenDyslexic mode is active, it takes priority over custom font selections.

Light & Dark App Theme

Use the toggle switch at the top of the app to switch between light and dark themes. Your preference is saved automatically and persists between sessions.

Dark Mode Output

Dark mode output inverts the colour scheme of your exported documents to display light text on a dark background. Benefits include:

• Reduced eye strain: Especially beneficial when reading at night or in low-light environments
• Less blue light: Dark backgrounds emit less blue light, which may help with sleep patterns
• Battery saving: On OLED screens, dark mode can significantly reduce battery consumption
• Reduced glare: White backgrounds can cause uncomfortable glare in dark environments

Dark mode is preserved when you export to HTML and EPUB formats.

🔧 Troubleshooting

Having issues? Find solutions to common problems below. If your issue isn't listed, contact us at [email protected].