MAI-Voice-2.1 vs Sume text to speech: cost per 1,000 characters

Sume text to speech costs $0.0475 per 1,000 characters, and a job of 1,000 characters bills 5 cents. Microsoft's launch post lists MAI-Voice-2.1 at $22 per million characters ($0.022 per 1,000) and MAI-Voice-2.1-Flash at $15 per million ($0.015), so on list price per character MAI is the cheaper of the two.

Where each number comes from

Sume's rate is in code: one Cartesia credit per character at $38 per million credits, times the 1.25 margin, which is $47.50 per million characters (CARTESIA_CREDIT_LIST_USD_MICROS_PER_CREDIT in packages/provider-pricing/src/pricing-tables.ts, public on the API pricing page). Each job is rounded up to a whole cent and capped at 20,000 characters.

The MAI figures come from Microsoft's launch post, read 2026-10-06; a third-party tracker lists the same numbers. Confirm them with Microsoft before budgeting.

What the comparison does not cover

This is a price-per-character table only. It says nothing about voice quality, languages or latency, and the tracker lists Flash with a latency claim that Sume's async jobs do not compare to. Short scripts show Sume's rounding most: 450 characters is 2.1 cents of rate but a 3 cent bill.

If you need voices that Sume lists, price them on Sume. If a different provider's voice is the requirement, this page cannot tell you which is better, only which is cheaper per character. For long scripts see 100,000 characters in five jobs.

• Sume: $0.0475 per 1,000 characters, 20,000 per job.
• MAI prices: from Microsoft's post.
• Round-up favors long jobs.

How to decide

Price per character is only one input. If your scripts are short, say 200 to 500 characters, both options cost a few cents or less per read and the difference rarely matters. At 20,000 characters per job, the list-price gap is $0.44 or $0.30 against 95 cents, so volume makes it a real number; at 1,000,000 characters it is $22 or $15 against $47.50 before Sume's per-job rounding.

Do not choose on this table alone. Compare voices on the text you will actually record, and confirm the availability and price of any provider directly. Sume's price is a code-verified quote; the MAI price is Microsoft's published list price, not a quote you can verify here.

How these numbers were produced

These figures come from Sume's own price estimators, run on 2026-10-06 through estimateCartesiaTtsProviderCost for Sume TTS, covering scripts of 450, 1,000, 5,000 and 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.