63,500-character chapter: Sume TTS $3.02 vs Aura-2 and ElevenLabs

Sume TTS charges $0.0475 per 1,000 characters, so a 63,500-character book chapter costs 63.5 x $0.0475 = $3.01625, which rounds to $3.02. The request limit is 20,000 characters, so the chapter goes out as four requests: three of 20,000 characters and one of 3,500. On the same text, Deepgram Aura-2 is $1.91, ElevenLabs v3 is $5.08 and ElevenLabs v4 Turbo is $2.54 at its regular price.

Rates were read on Sume's catalog, Deepgram's pricing page and ElevenLabs' API pricing page on 2026-10-09. ElevenLabs currently shows a promotional price for v4 and v4 Turbo that ends on October 12.

The comparison

All rows price the same 63,500 characters. Spaces and punctuation count on Sume, which is the usual convention for per-character billing, but check each vendor's own counting rule for your text.

Splitting a chapter into four requests

Because one Sume TTS request tops out at 20,000 characters, a chapter of 63,500 characters becomes 20,000 + 20,000 + 20,000 + 3,500. The largest request costs 20 x $0.0475 = $0.95, which is also the maximum the catalog shows for one request.

Cut at paragraph or sentence boundaries, not at exactly 20,000 characters, so the voice does not restart in the middle of a sentence. Each request returns its own audio, and Sume's timeline audio job can join the four files without gaps for a flat $0.01. Generated audio over 1,200 seconds fails with tts_duration_exceeded and captures no credit, so very slow speech settings can force even shorter requests.

What the cheap rows leave out

A lower rate only helps if the voice suits the book. Listen to one page from each candidate before committing 63,500 characters, and check how each handles names, numbers and dialogue. If you will need word timings later for captions or an audiobook sync, confirm that the option you pick returns them. Sume's TTS can return word timings and per-sentence slices when you ask for them in the request.

• Cost depends on the character count, not on the audio length.
• Four requests, one $0.01 concat: total $3.03 on Sume including the join.
• Re-recording a bad passage costs its characters again, so edit the text before you generate.

Budgeting a whole book

A chapter is a good unit for a budget. If the book has twelve chapters of about 63,500 characters, the Sume TTS total is 12 x $3.02, or about $36.2, plus twelve cents for joining pieces. On Deepgram Aura-2 the same book is about $22.9, and on ElevenLabs v3 it is about $61.0.

Those totals assume one pass with no re-recording. Plan for some passages to be redone after a read-through, and keep the text edits in your script file rather than in the audio so that a fix costs only the characters you change.