Tutorial 4: From document to published audiobook — voices, Read Aloud, and the Audio Studio

Goal: hear your document in the voice you want, then walk the Audio Studio's three journeys: narrate a folder of documents into a chaptered audiobook, combine a folder of recordings into one book, fix its chapters by ear, and publish the result — plus DAISY for accessible book players.

Everything here is keyboard-first and announced; a screen reader narrates every step, and nothing is applied without your review.

1. Pick a voice

Tools > Speech > Speech and Dictation... opens the speech hub. QUILL's voice lineup spans eras:

Every download is checksum-verified with visible progress.

2. Read aloud

Select text (or just place the caret) and use Tools > Reading & Dictation > Read Aloud. Stop on a key. Try the same paragraph in two or three engines — voices are a taste, and QUILL does not judge.

For fine control of a passage, the SSML Builder composes emphasis, pauses, and prosody from accessible controls — no hand-written tags — and plays natively on SAPI 5 and eSpeak NG.

For a quick one-shot file, Palette > Generate Speech Audio renders the current document to audio in the background. WAV always works; with the ffmpeg helper installed (Help > Download Optional Components), you also get MP3, M4A, M4B, OGG, Opus, FLAC.

3. Open the Audio Studio and pick a journey

For anything bigger than one file, Tools > Speech > Audio Studio... is the workshop (also in the command palette, default binding Ctrl+Shift+Grave, Y). The first page asks the only question that matters:

The wizard remembers your last journey and pre-selects it. Every page is announced ("Step 2 of 7: What should I read?"), Back and Next move between steps, and Skip to summary fast-forwards when a saved project profile already fills every page — a repeat run is three keystrokes.

4. Journey one: narrate documents

Point What should I read? at a folder of documents (Word, Markdown, HTML, text). Filters, subfolders, and a size cap keep discovery honest; Count documents speaks the settled number.

On Who should read it?:

On the output step, two boxes deserve attention:

On the book page, type a half-remembered title and press Look up book details — Open Library and MusicBrainz (free, keyless; QUILL asks before the first contact) fill the author, genre, and year from the match you pick, and offer to download the jacket as cover.jpg. Tick the spoken credits if you want the book to introduce itself.

Review the plain-sentence summary, optionally Save a job file (a portable .quilljob that pins the entire run — load it on the first page next time, or edit it in Notepad), and press Start. Progress is announced, per-file, cancelable, and minimizable to the status bar.

5. Journey two: combine a folder of recordings

Recorded a memoir chapter per session? Point the combine journey at the folder: each file becomes a chapter titled from its name, and you always review the chapter list before the merge — rename, reorder, remove, import titles from a CSV or Audacity labels. Output is a real M4B (native chapters) or MP3 (chapter markers), with optional silence trimming, fades, tempo, and ACX loudness.

Two scale-ups when you need them: Library mode builds every subfolder as its own audiobook, unattended, each titled after its folder; and the watch action ("Build audiobook from the folder") rebuilds a watched folder's master automatically whenever new recordings land.

6. Journey three: the Chapter Workbench

Open any chaptered MP3 or M4B — or a chapterless three-hour recording, which opens as one big chapter, ready to carve.

The built-in player anchors everything: Play/Pause, Previous/Next chapter, Rewind/Forward, a position slider that speaks human time, a pitch-preserved speed control, and Where am I? for the full audible glance. The player remembers where you stopped in every book. (Want gapless audio and exact seeking? Download the mpv player engine from Help > Download Optional Components and playback upgrades to libmpv automatically; QUILL falls back to the built-in engine whenever mpv is absent.)

The surgery, all by ear:

  1. Play until you hear where a boundary belongs.
  2. Split at playhead — the fix-a-bad-chapter move.
  3. Set start to playhead retimes an existing boundary; Merge into previous and Restore original round it out.

Two analysis helpers propose, never impose: Propose chapters from silences... scans the recording and lands a proposal in the list for review; Propose AI titles... slices each chapter's opening minute, transcribes it on your machine with the local speech model, sends only that text to your configured AI, and proposes a short title per chapter — review, rename, or Restore original. Check against ACX speaks a plain-words loudness verdict before you submit anywhere.

Saving is honest about physics: an MP3 saves in place (tags only, audio untouched); an M4B saves as a new file via a lossless re-mux.

7. Publish it

The Workbench's Publish... button offers explicit, consent-first paths:

Build the episode, regenerate the feed, upload both — the entire show runs from QUILL.

8. DAISY talking books

File > Export > DAISY Talking Book writes a DAISY 2.02 text-only talking book — the standard for accessible book players and library services. Your headings become the DAISY navigation structure, so build a clean heading outline first (GLOW will happily check it: tutorial 6).

9. Round trip: audio in

The same suite works in reverse — Ctrl+F9 dictates offline via Whisper, and transcription turns a recording into a document privately. Feed a transcript to the Listening Companion and get meeting minutes, action items, or a summary, reviewably. QUILL Cast episode 24 walks this whole tutorial by ear.

Next: Start an Accessible Vault.