QUILL — Meet You Where You Are

The screen-reader-first writing studio, built by the people who depend on it. This is the public beta, build 0.8.0 Beta 1.

From Community Access. Free. Optional by design. Private by default. Yours to make quiet.

"I started QUILL, but the community is lifting it to levels I could never reach alone. This release is about meeting people where they are, then giving them a path to grow." — Jeff Bishop

The announcement

For years, blind and low-vision writers have been asked to choose. Choose a plain editor your screen reader can read but that does nothing for you. Choose a powerful editor that does everything except work with your screen reader. Choose one tool to transcribe, another to proofread, another to make an audiobook, another to check a braille file — and choose, every single day, to fight software that was never built with you in mind.

QUILL is the end of that bargain.

QUILL — Quality, Usable, Inclusive, Lightweight, Literate — is a writing studio that puts the screen-reader user first. Not as an accessibility checkbox bolted on at the end, but as the design center everything else is built around. It writes, edits, and saves like a fast plain editor on day one. Then, only when you want them, it opens up an enormous range of capabilities — private on-device speech, document-to-audiobook production, braille proofreading, talking-book export, guided proofing, multilingual narration, and a whole writer's toolkit — every one of them optional, every one of them spoken, every one of them keyboard-first.

This document is the QUILL story and what you can do with this build today. This public beta — versioned 0.8.0 Beta 1 — puts the current feature set in your hands and the community's hands. It is a work in progress, and we will keep refining it, release after release, with your feedback driving what comes next.

An apology, and a commitment

Before anything else, we owe you an honest word.

The 0.7.0 release shipped with some fundamental issues, and it then took us far too long to get a better build into your hands. We are sorry. You trusted us with the tool you write with every day, and a long gap between a flawed build and a fixed one is exactly the kind of thing accessible software should never do to the people who depend on it.

We are changing how we work because of it. We are adjusting our release cycle to be more frequent and more responsive, so that a problem you report does not sit unanswered for days, and so that improvements reach you in smaller, steadier steps. We believe this version represents a much higher quality bar than 0.7.0, and we intend to keep raising it.

We are also not going to pretend everything is finished. We already know about areas that need more attention — the AI menu wants a cleaner structure, our Compare Mode features need more aggressive real-world testing, and the feature-entitlement process (which features show up for which profile) needs more polish. We will keep working on these in the open.

Again, we apologize. We hope you will understand that we are a small team, working and devoting tremendous time and effort into this project because we believe in it and we use it ourselves. Thank you for your patience, and for staying with us.

A thank-you contest — help us raise the bar

As part of that commitment to raising the quality bar, we want your help — and we want to reward it.

We are giving away five copies of Microsoft Office 365 Family Edition (valued at $129 each) to the top five contributors during this beta. We are looking for the people who provide thoughtful feedback, find and report issues, and show up as team players and community builders — the people who believe that building is done by everyone who picks up an oar to assist and drive the work forward with passion.

This is our way of saying thank you for your time, your patience, and your involvement. We value every one of you, and we wish we could do more — but we hope this sparks some interest in helping us reach a quality bar we will all be proud of.

How to take part:

What came before

QUILL did not start as one app. It started as a family of focused tools.

Across the accessibility world, small teams built tools that each did one hard thing well. One turned recordings into text and documents into speech, privately, on your own computer. ChapterForge assembled folders of audio into real, chaptered audiobooks. And a family of writing utilities — sound packs, compare mode, encoding repair, citation help, braille translation — grew up alongside them, the work of Blind Information Technology Solutions and CSE Designs.

Each was good. But for the person using them, the seams showed. Separate installers. Separate mental models. Separate places your work could live. Transcribe in one app, paste into another, proof in a third, and hope the formatting survived the trip. Every handoff was a place for a screen-reader user to get lost.

QUILL is the decision to stop shipping seams. We took what each of those tools did best, held it to one uncompromising standard — does it launch cleanly, does it speak clearly, does it work entirely from the keyboard, does it keep your work on your machine — and re-homed the survivors inside one editor. We left behind, or deferred to a later release, anything that could not yet meet the bar.

The result is a single, coherent home for writing the accessible way — and a foundation deliberately built to keep growing.

A community release, in the truest sense

The bigger story is not only what changed. It is who helped change it.

QUILL is community-built and community-shaped. People are testing with JAWS, NVDA, and Narrator. People are contributing code. People are filing issues, describing real-world workflows, stress-testing accessibility, challenging assumptions, improving prompts, writing documentation, and cheering the work on when it gets hard. All of it counts.

Some of that work has names on it. Kelly Ford contributed the Vision Prompt Library that gives image description twelve evaluated styles. Robert Danaraj contributed Math Equations, redesigned as a sandboxed Quillin. Taylor Arndt, Shane Popplestone, Michael Babcock, and so many others have tested builds, opened issues, and pushed QUILL to be better than any one person could make it. The braille tools exist because of the screen-reader users and braille proofreaders who filed the issues that drove them, and because of the maintainers of liblouis and the Universal BRF Pack.

"The future of QUILL is not one person deciding every feature. It is a community of people building the tools they wish existed, with accessibility baked into the foundation." — Jeff Bishop

To everyone who tried a build and told us where it hurt: thank you. This release is yours.

The promise behind every feature

Before the feature tour, the rules QUILL is built to. They are not marketing — they govern every line below.

"A feature is not finished when it exists. It is finished when the person using it feels confident, respected, and in control." — Jeff Bishop

Meet you where you are

QUILL does not force every user into the same cockpit. A first-time writer opens a quiet editor and starts typing. A braille professional inspects page, line, and cell with confidence. A developer moves through code by tokens. A reviewer compares files without scanning a diff. A power user extends the editor with Quillins. Everyone starts at the right level and grows from there.

A short startup wizard asks one question — what kind of writing do you do? — and shows a plain-English, screen-reader-readable preview of exactly what you will get, with no jargon and no list of what you will not get. Seven starting profiles, from Just a Text Editor (QUILL at its quietest) through Writer, Markdown and Web Author, Accessibility Professional, Braille Professional, AI-Powered Author, and Developer and Power User, are a curated front door to the full catalog of ten built-in feature profiles. Press Alt+Shift+P anytime to switch, reading the same "what you get" description before you commit. Menus, the Command Palette, Go to Anything, the system tray, and even Preferences all adjust instantly — commands for features you have not enabled simply do not appear. Cancel the wizard and you land in the simplest possible editor, never an overwhelming wall of defaults.

"Meeting people where they are means respecting beginners, professionals, power users, and explorers equally. QUILL should feel welcoming on day one and powerful on day one hundred." — Jeff Bishop

What follows is everything QUILL can do once you are ready for it.

1. Control what QUILL says — the Verbosity system

Most software has one volume of chattiness, and it is rarely yours. QUILL gives you a dial, from full friendly context for every action down to near silence, without ever fighting your screen reader.

2. Speak and listen, privately — offline speech

QUILL has a privacy-first speech engine that runs entirely on your computer, under Tools > Speech. No AI account, no key, no provider. The engine ships in the installer (the offline speech engine (whisper.cpp) component, included in a Full install) — no separate download, no PATH editing. Pick a model from Manage Speech Models and you are ready.

Manage models with confidence. The dialog opens with a read of your machine (RAM, whether a GPU was found), shows roughly how much memory each model needs, flags models too big for your RAM, marks the best fit as "Recommended for your computer," and warns when a large model has no GPU behind it. Downloads run in the background with a real, cancelable percentage and a low-disk warning. The models are open-source and MIT-licensed — no Hugging Face account, no license to click through, no gated files — and every download is checksum-pinned so a silent re-upload can never swap a model underneath you.

Transcribe audio and video, on-device. Turn a recording into text without uploading a byte. Output as plain text, Markdown, or HTML; it opens as an editable draft while QUILL works in the background. With the speaker-detection model installed, transcripts label who is speaking when. QUILL accepts a wide range of formats — MP3, M4A, AAC, FLAC, OGG, Opus, WMA, WAV, MP4, M4V, MOV, MKV, WebM, AVI — and can fetch ffmpeg for you (Tools > Speech > Download FFmpeg...) so you never have to prepare a WAV first.

Captions. Generate Captions (Offline) transcribes with timestamps and saves SRT or VTT subtitles ready to ship.

Transcribe a whole folder automatically. Point a Watch Folder profile at a directory, choose the offline Transcribe audio action, and every audio or video file you drop in is transcribed on your machine and saved beside it — in Text, SubRip, WebVTT, or Markdown. This automation never uploads, never uses a cloud provider, and tells you where to get a model if none is installed.

Cloud transcription, only if you ask. Add one as an extension — QUILL bundles three declarative Quillins: OpenAI Whisper (99+ languages), Groq Whisper (very fast turnaround), and ElevenLabs Scribe (high accuracy with optional speaker diarization). Audio is uploaded only when you explicitly transcribe a file with that provider — never offline, never in Safe Mode, never by the folder automation. The extension ships no code and never sees your audio or key; QUILL performs the upload through its audited network path.

Dictate into your document. Dictate (Offline) lets you speak straight into the editor — start, speak, stop, and QUILL inserts what you said as a single undoable edit. Then the two fast, keyboard-only ways that need no dialog and no focus change, on the same on-device engine:

Run commands by voice. Voice Command (Offline) drives QUILL hands-free — say "save", "bold", "next heading", "word count", "command palette". Recognition is on-device, and voice can invoke only a curated, safe set of commands, so a misheard phrase can never trigger anything destructive. Off by default, always disabled in Safe Mode.

Choose your engine. The bundled whisper.cpp needs nothing extra. On capable machines, opt into Faster Whisper (higher-throughput, multilingual, uses your GPU) or Vosk (very low-resource, CPU-only, ships in the installer for old or constrained machines). All three run entirely on your computer.

3. Read aloud and turn documents into audio

Read Aloud now spans local and cloud voices, and QUILL can produce real audiobooks from a folder of documents.

More voices, local and cloud. Local Read Aloud uses the Windows system voice through SAPI 5 directly, plus DECtalk (driven through its real synthesis runtime), eSpeak-NG, Piper, and Kokoro — and every catalog voice ships a short spoken preview so you can hear it before choosing. Kokoro voices now ship as part of the product, so its high-quality neural narration works fully offline out of the box, with no separate download. For cloud-grade narration, AI Voice supports OpenAI (11 voices), Google Gemini 2.5 (30 voices), and ElevenLabs (audiobook-grade export). QUILL shows an estimated cost before any export, splits long text only on sentence boundaries so audio never trails off mid-word, and adds a trailing pause so the last sentence is never clipped.

Documents to audio, in bulk. Batch Export to Speech Audio points QUILL at a folder, picks an engine, voice, and pace, and converts every document on a cancelable background task — with include/exclude filters and a maximum file size. Output as WAV, MP3, M4A, M4B (audiobook), Opus, FLAC, or OGG; stamp album, author/narrator, genre, and year; and turn long documents into chaptered audio with real MP3 chapter markers you can jump between in any podcast app, with a natural page-turn transition cue and configurable pauses.

Shape it. Export one chaptered file or a separate file per article; combine empty headings so you never hear hollow "chapter" announcements; round-robin voices so each article is read by the next voice in a list you build; normalize loudness to audiobook (ACX) level; and dry-run to write the exact spoken text for proofreading before paying for any synthesis.

Narrate in other languages. Pick one or more languages and a voice for each, and QUILL translates each document and narrates it — <doc> (Spanish).mp3 beside the original — for a whole folder or for just the open document (Export to Translated Speech Audio). Translation uses any AI provider you have configured or a local LibreTranslate; voices are local-first (eSpeak speaks nearly every language offline) with premium multilingual cloud voices as the quality tier. Pronunciation dictionaries can be scoped to a language, you see a combined cost estimate before any metered cloud run, and a project remembers its language targets.

Robust by default. If a voice fails to synthesize, QUILL remembers it and skips it next time, so one broken voice never derails a batch. When an output already exists, choose skip (cheap resume), overwrite, or rename. Mirror source folders or flatten them, rename with a template, and turn on a manifest for a CSV/JSON summary. A folder remembers its whole speech setup, so you configure a project once and re-run it anytime.

Teach QUILL how to say things. Manage Pronunciations adds corrections — names, brands, acronyms, technical terms — that apply everywhere speech happens, batch and live, with a live preview. Dictionaries can be global or per-project, and a starter dictionary ships with common terms covered. An optional text cleanup pass fixes the typography that trips up engines (curly quotes, dashes, ellipses, symbols, fractions, emoji) and speaks phone numbers, emails, and URLs clearly.

Fine pronunciation and prosody. A new SSML Builder composes emphasis, pauses, say-as, phonemes, and prosody from accessible controls, and Read Aloud plays that markup natively on SAPI 5 and eSpeak-NG, so the emphasis actually takes effect instead of being read aloud.

Build a full audiobook from a folder. Build Audiobook from Folder combines a folder of audio into a single chaptered MP3 or M4B master (native chapter atoms) with book tags and an auto-detected cover. Before building, rename, reorder, and merge chapters; then one-click Normalize to ACX brings the master into the loudness range audiobook platforms require, measured and reported. This is ChapterForge, re-homed inside your editor.

4. Read and proof braille

Braille support is not a novelty bolted on — it is designed around the way braille professionals actually work, and it never changes your braille file. Progress and notes live in a small companion file beside it.

"Braille users deserve tools that understand braille workflows, not tools that merely tolerate braille files." — Jeff Bishop

5. Save accessible talking books — DAISY

File > Export > DAISY Talking Book saves your document as a DAISY 2.02 text-only talking book — the format read by DAISY software and by hardware players like the Victor Reader Stream, Plextalk, and APH units. It exports what is on screen (no need to save first), your headings become the player's navigation points, and because it is a standard DAISY book you can open the folder later in a tool like APH Book Wizard Producer to add recorded or synthesized audio.

6. Proof your work — guided F7 Spelling Review

Press F7 and QUILL walks you through every misspelling in your document or your selection, one at a time. No AI, no network, nothing uploaded — the same local dictionary engine that powers as-you-type checking.

The heart of it is a read-only Context field that shows the misspelled word highlighted inside its sentence, where focus lands automatically; arrow through it character by character, move by word, copy — just like the editor, and Alt+W brings you back to the word whenever you need it. The actions are Change, Change All (capitalization preserved: teh→the, Teh→The, TEH→THE, as one undo step), Ignore Once, Ignore All, Add to Dictionary, and Undo Last — all by keyboard, all announced, with nothing double-read.

Tune everything under Settings > Spelling Review: announcement verbosity (Concise / Balanced / Detailed), spell-the-word-aloud letter by letter with an adjustable pause, sentence-or-paragraph context, and wrap-to-beginning. And Ctrl+F7 / Ctrl+Shift+F7 jump to the next or previous misspelling right in the editor without opening the dialog at all.

7. Write in any language — Document Language

QUILL now lets a document be edited as a language even when its filename says otherwise — so a plain .txt, an unsaved buffer, or a pasted snippet can get real HTML, Markdown, or code editing.

Set the language from the Format > Document Language submenu (or Navigate > Set Document Language...), and Bold/Italic insert the right markup, the heading/table/list/tag items light up, comment toggling uses that language's syntax, and the outline, link insertion, and live preview all follow it. Recognized languages include HTML, Markdown, CSS, Python, JavaScript, TypeScript, C, C++, C#, PHP, Go, Rust, Kotlin, Shell, YAML, JSON, TOML, and SQL. It is an editing aid, not a rename — setting HTML on a .txt reminds you to Save As. And optional auto-detection (off by default) recognizes code you paste into a plain file, in your choice of Hint, Suggest-and-announce, or Switch-automatically modes — cautious, never guessing on prose, never overriding a real extension, and unlike editors that switch silently, always keeping a screen-reader user informed.

8. Find your way faster — Quick Nav

Go to Anything (QUILL key + G) is one search panel for everything worth jumping to: headings, links, lists, tables, block quotes, bookmarks, code blocks — each with a live count and a type-ahead filter. It also lists your document's misspellings and, when a search is active, the current query's search hits as their own navigable types, so you can open one panel and jump straight to the next misspelled word or the next match the same way you jump to a heading.

9. Build lists by concept — Structured List Studio (F2)

Press F2 and build lists by working with plain concepts — item text, a checked box, a term and its definition — while QUILL writes the correct Markdown or HTML for you. You never type -, 1., [ ], <ul>, or <dl> by hand. Four list types (Bulleted, Numbered, Checklist, and Definition), live read-only source as you work, full nesting with subtree moves, multiple terms and definitions, and convert/import that warns before dropping structure. With text selected, F2 turns your selection into a list; with nothing selected it starts a new one; inside an existing list it loads that block back in to edit. Every control is keyboard-reachable and labelled, and inserting the finished list is a single undo step.

10. The writer's toolkit you already had — now in one place

QUILL also carries forward the toolkit that grew up across earlier releases — the durable value of those separate apps and earlier betas, re-homed on QUILL's invariants:

11. A smoother upgrade, a cleaner start, and a sturdier app

What happens when you update — and how we kept the impact small. If you have been testing an earlier beta, this build changes how QUILL remembers your settings so that future updates stop fighting you. Here is exactly what to expect, and the work we did to make sure none of it costs you anything:

  1. You install over your existing copy as usual. The installer now replaces QUILL's program files cleanly instead of layering new files on top of old ones. That mix of old and new files was the cause of the occasional crash some testers saw right after updating; it cannot happen anymore. To keep installs fast, only QUILL's own code is refreshed — the bundled Python runtime, voices, braille tables, and tools are left in place — and your documents and settings in %APPDATA%\Quill are never touched by the installer.
  2. The first launch tidies your settings once — after making a backup. QUILL converts your old settings file to a new, smaller format that stores only the choices you actually changed. Before it rewrites anything, it copies your original file into a migration-backups folder, so the change is always reversible. From now on, anything you customized is kept, brand-new options show up already set to a sensible default, and an improved default reaches you automatically unless you had changed that option yourself. This is the fix for the long-standing problem where a better default never reached people who had already used QUILL.
  3. A few important shortcuts are corrected once. Where a test build had moved something it should not have — most visibly Find, which returns to Ctrl+F — QUILL puts it right a single time, then never touches it again, so you stay free to rebind it. Prefer QUILL never change your shortcuts? Turn off Apply recommended keyboard-shortcut updates in Settings.
  4. Old crash noise is cleared away. The first time you launch the updated app, QUILL clears out the backlog of old log files and crash reports from earlier runs, so you start clean instead of staring at a pile of past problems. This happens once and only removes diagnostic files — never your documents, autosaves, backups, or recovery data.

Everything below is the detail behind those four steps.

How QUILL meets you where you are

Read those eleven sections again and notice what is not there: no demand. QUILL does not require you to learn a new way to write. It does not make you turn on AI. It does not move your files to a server. It does not talk over your screen reader.

A student writing a paper meets QUILL as a fast, faithful editor with citations and a table of contents. A podcaster meets it as a transcription and captioning studio. An author meets it as an audiobook pipeline. A braille transcriber meets it as a proofreading workflow that never touches the source. A blind developer meets it as a code-aware editor that finally announces the indent. A person who just wants quiet meets it as an editor that will, genuinely, be quiet.

That is the whole idea. One studio, many doors, and you choose which to open.

"Delight is not about fireworks. Delight is when the editor does the small helpful thing at exactly the right moment and then gets out of your way." — Jeff Bishop

What comes next

QUILL is a foundation, and it was built to grow. The roadmap is tracked openly, and every item here is optional and screen-reader-first by the same rules as everything above.

Some of these are close; some are further out. All are built on the same promises, so nothing new will ever cost you the quiet, the privacy, or the control you have today.

Availability

QUILL is free. This public beta includes:

We keep refining QUILL with live screen-reader use across JAWS, NVDA, and Narrator, and installer verification on clean Windows 10 and 11 machines. That is work the community does best, which is why the feature set is in your hands now.

How to help shape QUILL

The fastest way to influence QUILL is to use this beta for real work — write with it, review with it, customize it, push it. Then tell us what works, what breaks, and what should be better:

  1. Install or start QUILL.
  2. Let it carry forward your compatible settings at startup.
  3. Keep working with your familiar files and preferences.
  4. Send feedback from Help > Report a Bug — bugs, accessibility issues, confusing workflows, and feature requests all land where the team can act on them.

Optional extras, only if you want them: the offline speech engine ships in the installer (or drop it into tools\speech\whispercpp in a portable copy), then download a model from Manage Speech Models; ffmpeg (one click from Tools > Speech > Download FFmpeg...) lets you transcribe formats beyond WAV; and Faster Whisper adds a GPU engine on capable machines. None are required to launch, edit, or save.

A word on feedback — and a request

Please remember that this is a beta. Bugs will exist — that is the nature of a public beta, and finding them is exactly what this stage is for. We are trying very hard to stay on top of your feedback, and we will be more diligent in that process going forward. The improved crash reporting and the direct, in-app submission experience are built precisely to make that easier — for you to report and for us to act on it quickly.

And one request: be kind, and be constructive. Honest, frank feedback is welcome — we can handle that, and we genuinely want it. A small team reads every report, and a clear, generous one helps us fix the right thing faster.

About QUILL

QUILL — Quality, Usable, Inclusive, Lightweight, Literate — is a screen-reader-first writing studio from Community Access, built on the conviction that accessibility is not a feature you add at the end but the center you design around. It consolidates the durable value of a family of accessibility tools — including a private, on-device speech suite and the ChapterForge audiobook workflow — into one optional-by-design, private-by-default, keyboard-first environment. QUILL owns the editor, the focus, the undo, and the announcements; AI and every external integration are optional and off by default. Platform scope is Windows (primary) and macOS (supported).

QUILL is community-built and community-shaped. With QUILL, accessibility is not the finish line. Accessibility is where we begin.

Learn more at https://quillforall.org.