5,000 minutes of speech in 2027: Gemini TTS vs MAI-Voice vs Sume
At 25 audio tokens a second and 900 characters a minute, 5,000 minutes cost $135 on Gemini 3.8 Flash TTS in 2027, $99 on MAI-Voice-2.1 and $213.75 on Sume.

For 5,000 minutes of speech in 2027, Gemini 3.8 Flash TTS audio output lists at $135 and Flash-Lite at $90 once the promotional price doubles on 1 January 2027; MAI-Voice-2.1 costs $99 and its Flash model $67.50; Sume TTS costs $213.75. These use my assumptions of 25 audio tokens a second, which follows from Google's own per-10-second figures, and 900 characters a minute. Sume is the most expensive per minute here, and the only one of the four that sits inside a job API with captions, music and a render.
How each number is built
Gemini: 5,000 minutes x 60 s x 25 tokens = 7.5M tokens. At $18 per 1M that is $135, and at $12 it is $90. In 2026, at $9 and $6, the same volume is $67.50 and $45. Text input is extra: $0.50 per 1M now and $1.00 from 2027 (Google AI, read 2026-10-09).
MAI: 5,000 minutes x 900 characters = 4.5M characters. At $22 per 1M that is $99, and at $15 it is $67.50 (Microsoft AI, read 2026-10-09). Sume: 4,500 x $0.0475 = $213.75.
| Service | 2027 rate | 5,000 minutes | Per minute |
|---|---|---|---|
| Gemini 3.8 Flash TTS | $18 per 1M audio tokens | $135.00 | $0.027 |
| Gemini 3.8 Flash-Lite TTS | $12 per 1M audio tokens | $90.00 | $0.018 |
| MAI-Voice-2.1 | $22 per 1M characters | $99.00 | $0.0198 |
| MAI-Voice-2.1-Flash | $15 per 1M characters | $67.50 | $0.0135 |
| Sume TTS 1.0 | $0.0475 per 1,000 characters | $213.75 | $0.04275 |
What the numbers leave out
Gemini and MAI are model prices; neither includes captions, a render or an audio join. Sume lists none of the Gemini or MAI voice models. The Sume rate includes the job system, retries by idempotency key and the other media steps at their own prices. The assumption of 900 characters a minute is language-dependent, and Korean or Spanish scripts differ in length, so use your own measured figure.
A decision rule
If speech is your whole product and volume is high, the model-level prices decide, and you build the surrounding steps yourself. If speech is one part of a video pipeline, the extra $78.75 to $146.25 a year of 5,000 minutes against Gemini or MAI is small next to the render and clip costs. Recompute with real character counts and re-check each vendor page in December, because both Microsoft's and Google's rates are time-limited.
Sensitivity
The Gemini line depends on 25 tokens a second, which I derived from Google's own price per 10 seconds, and the MAI and Sume lines depend on 900 characters a minute. If your speech is denser, at 1,200 characters a minute, the character-priced lines rise by a third: MAI-Voice-2.1 to $132 and Sume to $285, while Gemini stays at $135. So the ranking of Gemini against the others depends on how many characters each minute of your audio holds.
Sources
Related posts
More in Pricing
- GPT Image 2.5 at 1024: xhigh is about 3,122 tokens and max about 7,024
Sume's docs put xhigh at $0.09366 and max at $0.21072 for a 1024 image at $30 per million tokens. That implies 3,122 and 7,024 tokens, before Sume's margin.
- GPT Image 2.5 quality auto holds the max price: a 12-image batch
Quality auto on gpt-image-2.5 (Flare, Sunburst) reserves max: $0.21072 of output at 1024x1024 before Sume pricing, so twelve images hold $3.16 or more.
- GPT Image 2.5 Batch is $15 vs $30 per M output tokens; Sume has none
OpenAI lists GPT Image 2.5 output at $30 per million tokens Standard, $15 Batch. A max 1024px image is $0.21 vs $0.11 in output tokens. Sume has no batch tier.
- GPT Image 2.5 low is $24.75 per 1,000: 25 cents under 2.5-cent rows
On Sume, GPT Image 2.5 at low 1K lists at $0.02475 per image, below the $0.025 of Grok Image, Imagen 4 Fast and Qwen Image. 1,000 drafts: $24.75 against $25.00.
Written by Sume