Sume text to speech price ladder: 500 to 20,000 characters

A 20,000-character script costs 95 cents on Sume text to speech, the most one job can cost because 20,000 characters is the cap. A 1,000-character script is 5 cents, 5,000 characters is 24 cents.

The formula

Sume charges one Cartesia credit per character at $38 per million credits, then adds 1.25: $0.0475 per 1,000 characters. The job total is rounded up to a whole cent (estimateCartesiaTtsProviderCost in packages/provider-pricing/src/media-estimators.ts). Public rates are on the API pricing page.

Splitting a long script

A 50,000-character script is two 20,000-character jobs and one of 10,000: 95 + 95 + 48 = 238 cents. Splitting at paragraph breaks lets you rerun one part without paying for the whole script, and the round-up lost across three jobs is under 2 cents.

The 20,000 limit is enforced on each job's text. The request fields are in the Sume API reference, and a worked comparison against another vendor is in the MAI price page.

• Cap 20,000 characters per job.
• Cost scales linearly except the round-up.
• Whole script at once costs the same as slices, give or take a cent.

Budgeting recurring narration

A weekly 3,000-character script costs 15 cents a run ($0.1425 rounds up), about $7.80 a year. A daily 800-character read is 4 cents, about $14.60 a year. At these prices the voice track is rarely the line that drives a media budget.

Count characters, not words or minutes. The billing unit is the character count of the text you send, so removing stage directions and markup before the request lowers the price directly. If you are unsure of the length, count it in code before submitting and split anything above 20,000 characters at a paragraph boundary.

How these numbers were produced

I read the rate tables in the clone of Sume's repository on 2026-10-06 and called estimateCartesiaTtsProviderCost for Sume TTS, covering scripts of 500 to 20,000 characters. The output fields are billableAmountUsdMicros and billableAmountUsdCents; the tables show the cents.

These are estimates of the billed amount for the stated settings. Anything not in the table, such as a different frame shape, a reference input or a changed quality tier, can change the number, so run the estimate for your own request before a large batch. Where this page and the public API pricing list disagree, the pricing page and the live quote win.