480,000-character audiobook: TTS cost, ElevenLabs v4 vs Sume

A 40-chapter audiobook of 480,000 characters (about 12,000 per chapter) costs $22.80 to narrate on Sume TTS. On ElevenLabs it is $10.56 with v4 at the sale rate and $38.40 at v4 list. The ElevenLabs API pricing page (read 2026-10-04) lists v4 at $0.022 per 1K characters until Oct 12, down from $0.08.

The decision is mostly about timing: before Oct 12 the sale price wins, after it Sume's $0.0475 per 1K is cheaper than ElevenLabs v4 list.

Why 40 jobs

Sume's TTS job accepts up to 20,000 characters (spaces and punctuation count), so a 12,000-character chapter fits in one job. Each chapter reserves 12 x $0.0475 = $0.57, and 40 chapters is 40 x $0.57 = $22.80. That equals the single-line total, so splitting costs nothing extra.

Chapter jobs are also easier to retry. If one chapter fails, only that job is refunded and resubmitted; the other 39 are untouched.

Joining chapters into one file

To get a single audiobook file, the Timeline 1.0 audio.parts[] input joins audio parts inside one render without re-synthesising speech. A render is listed at $0.10 per output minute, rounded up. If you join each chapter's audio into one 12-minute file, that is 12 x $0.10 = $1.20 per chapter, or 40 x $1.20 = $48.00 for the whole book, so only join where you need a single file.

• Chapter files as delivered: $0 extra.
• Joined chapter by chapter in Timeline: $1.20 each at 12 minutes of audio (an assumed length; measure yours).

Steps for a book-length run

• Split the manuscript at chapter breaks and count characters per chapter; flag any chapter near 20,000 characters and split it at a paragraph.
• Submit the chapters as separate TTS jobs and keep the returned audio URLs in chapter order.
• Sum the reserved amounts before you start: 480 x $0.0475 = $22.80 is your ceiling for the narration.
• Decide whether the listener needs one file. If not, ship chapter files; if so, join with a Timeline render.

If you revise chapters

Suppose four chapters (48,000 characters) are rewritten and re-recorded. That is 4 x $0.57 = $2.28 on Sume, 4 x 12 x $0.08 = $3.84 on ElevenLabs v4 at list, and 4 x 12 x $0.022 = $1.06 at the sale rate. Revisions are small next to the first pass, so choose the engine on voice, not on the re-record line.

Caveats

The 12,000-characters-per-chapter figure is an example to make the math concrete; use your real counts. Both vendors price characters, not words, so punctuation and markup count. Voice choice, pronunciation control and language coverage differ, so listen to a sample from each engine before you decide on price alone.

The sale rate ends Oct 12 per the vendor page. A job that starts after that date should be budgeted at the list rows.