An honest build log.
Written by the people doing the work, not by marketing. When something broke, it says so.
- Added
Rosters that grow with your series
Two roster upgrades for series work. First, you can now save a book’s cast back over the roster it came from: apply a roster, parse the next book, and one click updates the roster with the grown cast — new characters, recast voices, everything — without creating a duplicate. Second, verified castings travel with the roster: if you confirmed a narrator’s casting in Book 1, applying that roster to Book 2 keeps those characters verified, so their chapters skip the casting step instead of asking you to confirm the same voice again. The book page also now leads with the whole-book casting flow — parse, cast, then render — since casting once for the book is the natural next step after importing a manuscript.
- Improved
A UX audit, answered
An outside usability audit went through the whole product as an everyday author would, and this release closes out its findings. The big ones: book cover titles were rendering nearly invisible (dark ink on a dark jacket) and now show correctly; the terracotta button color was darkened one step so every call to action clears accessibility contrast; the trial meter now says plainly when credits have expired or are about to, instead of advertising an expired balance as spendable; and the simple pages — dashboard, bookshelf, billing, settings — finally work on a phone, with the sidebar collapsing into a menu. Casting got the biggest lift: choosing a voice now opens a real picker with search, gender and age filters, and one-click auditions, instead of a 40-option dropdown. Plus dozens of smaller things: clearer credit-history wording, honest review progress in the chapter stepper, one-click emotion tags, a growing performance-note field, browser tabs named after your chapters, an explicit “Mark book complete,” and a 404 page that recognizes you’re signed in.
- Added
ACX delivery: separate files per chapter
ACX-style delivery expects one file per chapter or section, not a single merged audiobook — so the exporter now does exactly that. A "Keep chapters as separate files" checkbox (on by default with the ACX-style preset) masters every chapter individually — loudness and room tone are applied per file, and every file is measured against the spec, including the 120-minute per-file limit — and delivers a ZIP of numbered MP3s with a readiness report inside. The export history shows a pass/fail check and a download link for each file. Mastered exports are also now pinned to 44.1 kHz, matching the ACX spec exactly.
- Improved
Takes keep their names
The tuning panel now shows the current take by name — "Take 2 · Current" — above the previous takes. Restoring swaps the labels instead of renumbering: restore Take 1 and Take 2 moves down into previous takes, ready to bring back. New takes always get the next number, so "Take 3" still means the third read you recorded.
- Improved
Studio quality-of-life: undo, honest plans, and safer renders
Deleting a line or a character now offers a few seconds of Undo instead of being instantly permanent. The Render button disables itself when a render would exceed your balance and points at top-ups, rather than letting the render fail. Playing lines through with Play All now marks them reviewed automatically. Small screens get a friendly "works best on desktop" notice instead of a broken layout, dialogs keep keyboard focus inside them, and going offline shows a banner instead of failing silently. The pricing page also got more honest: every plan is the same studio — tiers differ only in monthly hours, and we removed feature bullets that suggested otherwise.
- Improved
Terms, Privacy, and AI policies finalized
The legal pages are no longer marked as drafts pending review — the “plain-English template” notice is gone, and the Terms now name their governing law (the State of Idaho, United States). Because the terms changed, you’ll be asked to re-accept them once the next time you open the studio.
- Fixed
Honest top-up pricing, and two small fixes
Top-up copy now reflects the bundle rates — one hour is $10, but the 5- and 10-hour bundles work out to $9 and $8 per finished hour, and the pricing calculator quotes the cheapest bundle mix instead of a flat rate. Also fixed: the Skip tutorial button could ignore clicks when a tutorial popover overlapped that corner of the screen, and the import buttons now mention Markdown (.md) files, which have been supported all along.
- Added
A casting step, and voice verification that covers dialogue
Every parse now lands on a casting page before the script: see who’s new to the production vs. returning, map a duplicate name onto its existing character in one click, and audition any voice on that character’s actual lines — free, as many takes as you like. Auditions also measure each take’s pitch, so continuing locks a per-character verification range that’s tighter than the stock one. And linked two-speaker passages — previously unverifiable — now get a register check against the pair’s combined range, so dialogue-heavy chapters finally show honest voice-verified coverage instead of a wall of “not checked.”
- Fixed
Book details dialog could silently fail to save
On books without an uploaded cover, saving Book details hit an error under the hood and the dialog stayed open with no explanation. Saving works now, and if anything does go wrong the dialog says so instead of going quiet.
- Added
Publishing readiness checklists
Each book page now ends with per-store publishing checklists — Spotify, Google Play, Kobo, and the aggregators that accept digitally narrated audiobooks, plus an honest card explaining why ACX doesn’t and what Amazon’s KDP Virtual Voice alternative is. Automatic checks read your latest export (format, bitrate, loudness report, metadata); the rest are checkboxes that persist per book, with copyable disclosure lines and links to each store’s current policy.
- Added
A character bible for every book
The Cast section of the book page now shows the full company: each character’s voice with an audition button, age and register, aliases, and private continuity notes that never touch the audio. Aliases feed the director AI, so when the manuscript says “Ed” the studio maps it back to Edward instead of inventing a second character — and re-parses now recognize your existing cast even without a roster. Newly parsed chapters also record who speaks in them, so each card shows where its character appears. Aliases and notes travel with cast exports and rosters.
- Added
Ten new voices: children, teens, and elders
The voice library grew from thirty to forty. The new voices cover registers the catalog was missing — boys and girls of about eight, teenagers, and elders — each with its own audition sample. Voice pickers now group the library by Adults, Teens, Children, and Elders so the right register is easy to find.
- Improved
Every take is checked against its voice
The studio now verifies that each rendered line actually sounds like the voice you cast — same register, same character — and automatically re-records any take that drifts before you ever hear it. This closes the rare but jarring case where one line of a character came out sounding like a different person. Renders also retry transient engine failures on their own, which eliminated the intermittent line errors some chapters hit.
- Added
Roll a new take of any line
Re-rendering a line or passage now records a genuinely new take instead of reproducing the identical one, so if a delivery isn’t sitting right you can roll again until it does — the last five takes are kept for comparison and rollback. Previews and renders are now take-consistent too: the preview you approve is exactly what the render produces.
- Improved
Richer performance direction from the director AI
We overhauled how the studio directs the voice engine, following the current best-practice guidance for our TTS model. Parsing now writes concrete, sensory scene descriptions, per-character context, and performance notes drawn from the delivery cues the engine performs best. The practical result is more expressive, more consistent narration from the same manuscripts.
- Added
Play from any line, and add lines at the ends
Every rendered line now has a second play button that plays from that line to the end of the chapter, with the same pauses the finished chapter will have. You can also add new lines above the first line and after the last one — the end-of-script “Add a line” button is always visible now instead of appearing only on hover.
- Added
Download every take as its own file
A new download option saves each rendered line as its own numbered mp3, named for the character who speaks it, in script order — the raw takes exactly as they sound, with no assembly. Handy for mixing in your own tools or spot-checking performances offline.
- Improved
Voices stay truer to the one you cast
AI voices now hold their character more steadily through a render, and re-rendering a line reproduces the same take instead of rolling a fresh-sounding voice. We also rebuilt the audition clips so the voice you hear while casting matches what you get when you render.
- Added
Reorder and delete chapters
You can now drag chapters into a new order on the book page and delete a chapter you no longer need. Deleting a chapter that already has a script or rendered audio asks you to confirm first, and the remaining chapters renumber automatically.
- Improved
Editing the script got smoother
Inserting a new line now opens its text box with the cursor ready, so you can type what it says right away and pick the speaker from the dropdown. And when a single line fails to render, the studio suggests what to try based on that line — linking a very short one to a neighbor, or splitting a very long one — instead of a generic retry message.
- Improved
Start a new book by importing a manuscript
The New project button now offers to import a manuscript straight away — Word, PDF, EPUB, Markdown, or plain text — so you don’t have to create an empty book first.
- Added
Carry a cast between books as a file
You can export a book’s cast to a file and import it into another book or account — handy for a team producing a series. It uses the same format as cast rosters.
- Fixed
Parsing keeps narration like “he said”
The director AI no longer drops dialogue tags such as “he said” or “she whispered” when it scripts a chapter. Every word of your manuscript is kept and read by the narrator.
- Added
Import EPUB files
The manuscript importer now reads EPUB files alongside Word, PDF, Markdown, and plain text. It follows the book’s reading order from the EPUB spine and detects chapters automatically, the same as any other upload.
- Added
A guided tour on your first visit
The first time you open the studio, an interactive walkthrough sets up a short sample book and guides you through the whole workflow: pasting a manuscript, casting voices, previewing and re-previewing lines, rendering, and exporting. It runs the real pipeline for you without spending any credits, keeps the controls focused on one step at a time so nothing gets lost, and can be skipped whenever you like. You can replay it any time from Settings.
- Improved
Jump straight to a render error
When a line fails to render, the studio now shows how many lines errored and gives you a button to jump to the next one. It selects the line and scrolls it into view, so you can fix and re-render without hunting through the chapter.
- Improved
A roomier default pause between lines
New chapters now default to a half-second pause between lines instead of a quarter-second, which reads more naturally for most narration. You can still set the chapter-wide pause and override it on any individual line.
- Added
Plans and top-ups are live
You can now subscribe to a plan or buy credit top-ups right in the studio. Three monthly plans (Indie, Creator, and Professional) and one-time top-up bundles are available, with secure checkout and a self-serve billing portal for your subscription, payment method, and invoices. Credits land in your balance the moment a payment clears. The free trial still gives you an hour of finished audio to start.
- Added
Cast rosters keep a series consistent
Save a book’s cast as a reusable roster and apply it to the next book in the series, so recurring characters keep the same voice. When the director AI parses new chapters, it recognizes returning characters by name or alias and casts them automatically instead of starting from scratch. Manage your rosters from the new Cast Rosters page.
- Added
A Help Center for every step
Step-by-step guides now ship with the studio, covering each screen in order: creating a book, importing a manuscript, casting voices, directing lines, pronunciations, rendering and credits, review, and exporting an ACX-ready file, plus a billing and FAQ reference. Reach it from the site navigation and footer, or from inside the studio, where it opens in a new tab so you keep your place while you work.
- Added
The whole pipeline, working end to end
Render now reserves and settles credits honestly (failed lines are never charged), previews play real audio, chapter masters assemble automatically, and book export produces a chaptered M4B you can actually download. The studio gained line editing, speaker reassignment, re-parse, autosave, per-line playback, and a real progress bar. Also new: this page, the rest of the public site, and legal policies.
- Fixed
Manuscript parsing fixed
The parse step was calling a model alias that does not exist on our AI platform, so every manuscript failed with a server error. Parsing now runs on a stable GA model, and we completed a full audit of the production app to catch anything else pretending to work.
- Fixed
Login repaired, render pipeline wired
Google sign-in works end to end again, the AI render pipeline is connected to the studio, and deploys now follow a written runbook instead of tribal memory.
- Infrastructure
Platform hardening
Next.js security patches from the May advisories, Node 24 across the stack, and a stabler Cloud Run deployment after a week of wrestling hosting adapters.
- Added
First brick
Initial release: the marketing site, Google-only sign-in with an hour of trial credits, the studio scaffold, and a thirty-voice library with auditions.