One 450-character ad read priced five ways: 0.45 to 2.14 cents

One 450-character, 30-second read costs 0.45 cents on Gemini 3.8 Flash-Lite TTS, 0.675 cents on Gemini 3.8 Flash TTS and on MAI-Voice-2.1-Flash, 0.99 cents on MAI-Voice-2.1 and 2.14 cents on Sume TTS, at list prices read 2026-10-09. The cheapest engines are not on Sume. Sume's price includes its own 1.25 multiplier on provider cost, a single balance and the same API as the video steps that follow.

How each number is built

Microsoft's model page lists MAI-Voice-2.1 at $22 per 1M characters and the Flash version at $15 per 1M. Google's pricing page lists Gemini 3.8 Flash TTS output at $9.00 per 1M audio tokens and Flash-Lite at $6.00, through 2026-12-31, with 25 audio tokens per second. Sume lists $0.0475 per 1,000 characters.

A 30-second read is 750 audio tokens on Gemini. The 450 characters is a 15-characters-per-second assumption; it is a round pace for English, not a Sume or Google figure, so the two families of price (per character and per second) line up only as far as that assumption holds. Gemini text input is excluded.

What changes on 2027-01-01

Google's page doubles every Gemini 3.8 TTS rate on that date, so the Flash read becomes $0.0135 and the Flash-Lite read $0.009, and 100 reads cost $1.35 and $0.90. The Microsoft page read here prices MAI-Voice-2.1 and Flash without an end date; the cheaper launch rate that applies to MAI transcription is a different product.

When the price gap is the wrong thing to optimise

Four reasons the cheapest row may not be the right one.

• For 100 ads the whole range is $0.45 to $2.14. A single failed render or an hour of human review costs more than the difference.
• An ad read is the start of a chain. A 30-second vertical video on Sume adds a Timeline render at $0.10 and, if you want captions burned in, $0.20, so the voice is the smallest line.
• Voice fit decides which engine you can use at all. A language, an accent or a voice your brand already owns outweighs 1.5 cents.
• If you only need audio files for another editor, buy the cheapest engine that sounds right and skip Sume for that step.

Where Sume is the practical choice

Sume does not list Gemini TTS or MAI voices; its TTS catalog is the sume/tts-1.0 surface and the TTS Router ids sonic-3.6, sonic-3.5, sonic-3, sonic-latest and sonic-preview, all billed at the same per-character price. Use Sume when the audio feeds a Sume render, caption or avatar job and you want one key, one balance and one job record.