Gemini 3.8 TTS, 500 12-second reads: $2.43 priority, $0.68 batch

Five hundred 12-second reads is 500 x 12 x 25 = 150,000 audio tokens on Gemini 3.8 Flash TTS. At the 2026 standard output rate of $9.00 per 1M that is $1.35. Priority at 1.8 times the standard rate is $2.43; Batch or Flex at half the standard rate is $0.675. The same reads on Sume TTS, at an assumed 15 characters per second, are about $4.28.

What Google's page says

Google's pricing page, read 2026-10-09, states that audio tokens correspond to 25 tokens per second of audio. For Flash TTS it lists paid standard output at $9.00 per 1M through 2026-12-31 and $18.00 from 2027-01-01, Batch and Flex at half the standard paid rates, and Priority at 1.8 times standard. For Flash-Lite TTS the output rate is $6.00 per 1M through 2026-12-31 and $12.00 from 2027-01-01.

The page also lists text input at $0.50 per 1M through 2026 for Flash TTS. A 12-second read has a short script, so input is a small add-on and is excluded from every figure below.

The same job on Sume

Sume TTS charges $0.0475 per 1,000 characters. If a 12-second read is about 180 characters (an assumption of 15 characters per second), one read is $0.00855, and 500 reads are $4.275. Each read is its own job, and the catalog says the reservation rounds up to whole cents, so expect the reserved amount to run higher than the captured one for such short lines. Sume does not list Gemini TTS in its catalog, so this is a price check, not an option inside Sume.

Counting what is not in these totals

Every number above is output audio only. Gemini bills text input separately, and Sume bills per character with no separate input line, so the two are not perfectly aligned. Neither figure includes any retry. If 5 percent of reads need a second take, add 5 percent: that is $0.07 on the Gemini standard job and $0.21 on the Sume job. A read that is rejected or fails is not the same as one you dislike; the first is refunded by Sume, per the catalog, and the second is a new generation.

Choosing a tier

Four rules for a bulk read job.

• Priority is for a person waiting on the audio. Five hundred reads for a catalog update rarely need that, and it costs $1.08 more than standard here.
• Batch and Flex halve the price. The page excerpt read here does not state a turnaround time, so test one batch before you plan a deadline around it.
• A change on 2027-01-01 doubles every tier, so the same job is $2.70 standard and $1.35 batch in 2027. If the job is annual, schedule it before the date or budget for the doubled rate.
• If the reads become video, Sume's Timeline render is $0.10 per output minute. Six thousand seconds of audio is 100 minutes, which is $10.00 of render, and it exceeds the 1,800-second spine limit, so it needs at least four renders.