Table Read: directed multi-voice audiobook drafts for indie authors
Web app turning a manuscript into a directed, multi-character audio performance. Indie authors pay per finished hour.
Crowded · Open directable speech model, 3B active
Written for: Solo builder
Next step
Check the Fish Audio Drama 3 licence, then render one public-domain short-story chapter with three directed voices on a rented GPU and post a comparison to plain text-to-speech in r/selfpublish.
Why now
Older text-to-speech read flatly. Taking a director's notes and re-performing a line is the part human narrators were paid for. With 3B parameters active per token, a rented GPU can render a novel for a few dollars, so per-hour pricing well below human narration is viable.
What became possible
Fish Audio Drama 3 is an open-source speech model that follows natural-language direction. It revises a take when given a note, voices multi-character scenes, edits by infilling, and handles long-form audio. It uses a 35B mixture-of-experts backbone with only 3B parameters active per token.
Two ways in
- Table Read
Upload a manuscript, assign voices to characters, give notes per scene, and export chapter files.
- Customer
- Self-published fiction authors on KDP
- Model
- $15 to $30 per finished hour
- Serial-fiction podcast mode
Weekly episodes from web-serial chapters with recurring cast voices.
- Customer
- Royal Road and Substack fiction writers
- Model
- $12 a month subscription
Competition
The main risk is that ElevenLabs or Spoken adds direction notes or an episodic mode, so any edge would need to come from a better directing workflow and close ties with serial-fiction communities rather than from better voices.
Full competition notes (9 competitors)
Turning a manuscript into a multi-voice audiobook is no longer a new idea. Spoken does exactly this for indie authors, and ElevenLabs now detects and casts characters automatically and hands authors a route to Spotify and ElevenReader. Amazon and Audible own the free, single-narrator default for KDP authors. Competing on "multi-voice" alone would put you up against ElevenLabs, which owns the underlying technology. The most realistic opening is the "directed" part, together with pricing. The tools I found handle casting and fixing single lines, but none clearly offers scene-level direction notes that change how a whole scene is performed, with the voices staying consistent from scene to scene. ElevenLabs charges credits every time a line is regenerated, so authors pay for trial and error. A flat price per finished hour with revisions included takes that risk off the author. That price model is only a modest edge, since at least one service (AudiobookOne) already charges per finished hour, though it doesn't do multi-voice. The serial-podcast mode is less crowded on the open web. Pocket FM does AI serial audio but keeps it inside its own app, and a Royal Road author who made a full-cast podcast in 2023 described doing it by hand: about $500 in software and 2 to 4 hours of editing per chapter. A tool that keeps one recurring cast across weekly episodes and publishes straight to the author's own podcast feed or Substack looks like an open niche. Royal Road and Substack writers seem to have nothing purpose-built for this. The main risk is that ElevenLabs or Spoken adds direction notes or an episodic mode, so any edge would need to come from a better directing workflow and close ties with serial-fiction communities rather than from better voices.
- Spoken
The closest direct competitor. It is built for indie authors and turns manuscripts into multi-cast AI audiobooks, with voices designed for each character (one title has over 100). It runs on Hume's expressive TTS and promotes itself to authors through industry summits such as Expand Summit 2026, citing a company-promoted Edison Research study on listener preference.
- ElevenLabs Studio / ElevenCreative Audiobooks (Character Casting)
The biggest platform threat. Its Character Casting feature detects every character and suggests voices by age, gender, tone and role. Studio also supports manual per-section voice assignment, chapter projects, regenerating individual fragments, and publishing to ElevenReader and Spotify. It is sold as a credit subscription, roughly $11 to $330 a month, and reviewers say each regeneration uses extra credits.
- ElevenLabs for Spotify authors
A distribution partnership that markets character voices staying consistent across a whole script for authors publishing audiobooks to Spotify.
- Castly: Full-Cast Audiobook
A consumer app that automatically gives every character its own AI voice. It accepts uploaded files and fanfic links, so it is aimed at listeners more than authors.
- Amazon KDP Virtual Voice
Free single-narrator AI audiobooks for KDP authors, with a 40% royalty and list prices of $3.99 to $14.99. Authors report it can't voice characters differently. It is the default free option the product would have to beat for KDP authors.
- Audible AI narration for publishers
Launched in May 2025 with more than 100 AI voices, a managed service, a self-service option, and clones of narrators' voices. Per-character multi-voice support was not confirmed in what I found.
- Pocket FM AI Audio Series (with ElevenLabs)
Lets serial writers turn their stories into episodic audio series with AI voices and background music inside the Pocket FM app. It reports over 30,000 hours produced. It overlaps directly with the serial-fiction mode, but the audio stays on Pocket FM's platform.
- NarrationBox
An AI audiobook narration service aimed at self-published authors on subscription plans. It competes on price for the same buyer.
- AudiobookOne (Walnut Publication)
Sells AI audiobook production at £1 per minute, which works out to £60 per finished hour. Its per-finished-hour pricing is already in the market.
Checked 2026-10-14
Risks
- The model's licence may bar commercial use. Check first.
- Audiobook retailers' policies on AI narration may limit where finished files can be sold
- Crowded against ElevenLabs and similar hosted voice tools
- Voice-cloning consent abuse
Best on team fit; open space is the weak point (4).
tech 6 · pull 6 · space 4 · fit 7 · build 6
Exposed
Companies that could benefit or be threatened. News and context, not investment advice.
Momentum of its main topic
Sources
Status history
- 2026-10-14 Flagged as buildable