Spoken vs. Virtual Voice vs. ElevenLabs vs. Lex: Which AI Audiobook Route Fits Your Book? (2026)

Three names come up in almost every conversation we have with authors about AI audiobooks: Spoken, Amazon's Virtual Voice, and ElevenLabs. The fourth is always ACX, because a human narrator on Audible is still the mental default. This post puts them next to Lex on the questions that actually decide the outcome: what you pay, who does the work, whether you walk away with files, and where the finished book can sell.

The useful way to read the market is not five brands but three product shapes. Checkbox tools (Virtual Voice, and the Google and Apple equivalents) generate audio from your ebook and keep it inside one store. DIY tools (ElevenLabs Studio, Spoken, Speechify) give you an engine and a studio; you are the producer. Done-for-you production (Lex) casts, directs, scores, masters, and hands you retail-ready files. ACX sits off to the side as the human-narrator route with its own rules.

Key facts, as of September 2026: Spoken charges per 5,000 words, with a subscription that halves the rate, and you pay only when you publish. Virtual Voice is free, invite-only, US-only, pays 40% and never gives you an audio file. ElevenLabs is free to stream inside its own app or a monthly credit plan if you want to edit and export. ACX still requires human narration, still runs a 7-year exclusive term, and does not accept externally produced AI audio. Lex is a flat per-book production with the files, rights, and distribution choices staying yours; current packages are on lex.audio.

How do the five options compare side by side?

Here is the whole market on one table. Competitor figures are from each company's published pricing or help pages as of September 2026 and will move; the links go to the source.

	Spoken	KDP Virtual Voice	ElevenLabs	ACX	Lex
What it is	Self-serve AI studio	Amazon's built-in AI narration	TTS engine plus Studio editor; ElevenReader app for streaming	Human narrator marketplace	Done-for-you scored production
Cost model	Per 5,000 words, half price on a monthly subscription, pay at publish	Free	Free in-app streaming; monthly credit plans to edit and export	Per finished hour upfront, or Royalty Share	Flat per-book packages, see current pricing
Who does the work	You (casting, re-narration, edits)	Amazon, one automated pass	You (chunking, pronunciation, seams, mastering)	The narrator, over weeks	Lex (casting, direction, score, mix, QC)
Voices and cast	Single, dual, or multi-cast from a catalog or generated voices	Stock voices, one per chapter	Large library, cloning, multi-speaker in Studio	One human narrator per contract	Cast per book, optional multi-voice, author approves samples
Music and sound design	Not offered	None	Not in the audiobook path	Narrator-dependent, usually none	Licensed music and SFX on scored tiers
Files you own	Yes, MP3 downloads	No export	Streaming only on the free path; MP3/WAV export on paid plans	Audible holds distribution; you get the files	Yes, mastered chapter files and credits
Where it sells	Anywhere you upload	Amazon and Audible only	ElevenReader app, Spotify, or wherever you upload exports	Audible, Amazon, iTunes	Spotify, Apple, Google, Kobo, libraries, and Audible via publisher agreement
Exclusivity	None	Tied to your KDP ebook and Amazon	None	7-year term on exclusive deals	None
Turnaround	Minutes per chapter, plus your editing time	Minutes, up to 72 hours to go live	Minutes per chunk, plus your editing time	Weeks to months	Free first chapter in about 24 hours, full book in days

What is Spoken, and who is it right for?

Spoken is a self-serve AI audiobook studio that charges by word count and only bills you when you publish. You upload an EPUB or manuscript, it parses chapters, recommends or generates voices for the narrator and each character, and you refine chapter by chapter inside the browser. Its pricing page quotes a flat rate per 5,000 words for non-subscribers, half that on a monthly subscription, with unlimited re-narration and voice swaps included, and the finished MP3s are yours to publish anywhere.

Where it shines: prolific authors who like being in the driver's seat, and anyone who wants a multi-cast read without paying per voice. Where it stops: you are the producer. Casting decisions, pronunciation fixes, pacing, and listening back to every chapter are your hours, and there is no music or sound design layer. If you want the audiobook to arrive finished, Spoken is a very good tool handed to you, not a studio working for you.

What is Amazon Virtual Voice, and what does it cost you?

Virtual Voice is free to make and expensive to own nothing of. It is an invite-only KDP beta for US authors: pick one of Amazon's stock voices, set a list price inside Amazon's allowed band, and the audiobook appears on Amazon and Audible within about three days. Amazon pays a 40% royalty, the listing is clearly labeled as virtual voice, and your ebook cannot already have an Audible edition.

The two things it never gives you are a file and a choice of store. The audio lives inside Amazon's system; you cannot download it, put it on Spotify or Apple, or sell it from your own site. It is one automated pass with no casting, no direction, and no pronunciation review, and listeners can hear that. We do not think Virtual Voice is a scam; it is a reasonable zero-cost way to test whether a title has any audio demand on Amazon. It is just not an asset. If the test says yes, you will want a produced edition you actually own, and the Virtual Voice edition will have to come down first.

Is ElevenLabs a way to make an audiobook, or a tool to make one with?

ElevenLabs is the engine a lot of the industry runs on, including parts of our own pipeline, and it sells two very different things to authors. ElevenReader Publishing is free: upload your book and readers can stream it, on demand, inside the ElevenReader app. You cannot pick the voice in advance, edit the read, or export the audio. ElevenLabs Studio is the paid path: monthly credit plans (Creator and Pro are the tiers most authors land on) that let you assign voices, tune pronunciation and pacing, and export MP3 or WAV files for Spotify or wherever else you upload.

We will say the quiet part plainly because it is also why Lex exists: the engine is not the audiobook. A ten-hour novel through Studio is hundreds of generations you have to listen to, a voice that drifts in energy and pitch between takes unless someone manages continuity, pronunciation dictionaries you build yourself, and store specs (silence padding, loudness, sample length, credits tracks) that you learn by being rejected. Studio is a serious tool for authors who want to be their own producer. It does not cast, direct, score, or QC anything. Lex uses professional engines too; what you are paying for is the production around them.

Where does ACX fit in 2026?

ACX is still the only self-serve route to a human-narrated audiobook on Audible, and its terms have not gotten friendlier. You either pay a narrator per finished hour upfront or trade half your royalties in a Royalty Share deal, and either way the exclusive tier binds the title to Audible, Amazon, and iTunes for seven years. Audible is also moving royalties to a pooled model during 2026, so the per-sale math on the pricing page is a moving target. ACX does not accept externally produced AI narration at all, so none of the other four routes flow through it.

Pick ACX when the performance is the product: a memoir in a particular voice, a celebrity reader, a title with proven audio demand that justifies a four-figure narrator fee. The full side-by-side, including the royalty share math, is on our ACX alternative page.

What about Google, Apple, Speechify, and Spotify?

Google Play Books auto-narration and Apple Books digital narration are the Virtual Voice pattern in two other stores: free, generated from your ebook with the store's own voices, and locked to that store with no file to take elsewhere. Fine as a floor, not a performance, and not an asset.

Speechify Studio and similar subscriptions are DIY tools in the ElevenLabs Studio mold, often faster and built for high-volume voiceover. Same trade: you get an engine and an editor, and you do the production.

Spotify is not a production tool; it is where the finished book goes. Spotify for Authors accepts AI-narrated audio you produced yourself with a digital-voice disclosure flag, which makes it the friendliest major store for any of the export-capable routes above. Lex delivers there as part of the Publisher package. Store-by-store acceptance rules are in our distribution guide.

What does Lex do that a tool cannot?

Lex is the done-for-you column, and the work is the difference. Every production starts with a producer reading the manuscript and casting a narrator (or a cast) for that book, then sending the author samples to approve. Narration runs through our PROSE pipeline, which is built to keep one performance continuous across a ten-hour read: the same energy at hour six as at minute one, acted dialogue instead of read dialogue, and seam-by-seam QA so a chapter never jumps register mid-scene. Names and invented words get a pronunciation pass before the author hears them.

Scored tiers add licensed music and sound design placed where the text earns it, mixed under the voice rather than on top of it. Every file ships already passing retailer QC (head and tail silence, loudness, peak, retail sample length, credits tracks) because store rejections are where DIY projects go to die. Authors review everything in a private delivery room with a Follow along transcript synced to the audio, request line fixes, and download mastered masters they own outright, with a commercial license included and no exclusivity anywhere. The first chapter is free and lands in about 24 hours.

Which route should you pick?

• You want a zero-cost Amazon test and nothing else: Virtual Voice, eyes open about owning nothing.
• You want to produce it yourself and enjoy the studio work: Spoken if you want multi-cast and pay-per-word; ElevenLabs Studio if you want the deepest voice controls and export.
• You want a specific human performance and one store matters most: ACX, with the seven-year term priced in.
• You want a finished, scored, store-ready audiobook you own, without producing it yourself: Lex.

The fastest way to compare is with your own book. Send the manuscript and hear the first chapter performed, free, at lex.audio/first-chapter. If you would rather run the distribution side yourself afterward, the DIY playbook is public too.