Google Gemini TTS
Google · tts · reviewed by Stan K. · Source: data snapshot · data as of 2026-11-12
≈$0.014/min with the only published token rate in the table — 25 tokens/sec makes the math checkable, but the promo price doubles after 2026-12-31.
#5 of 16 in Tts
ToolScore breakdown — why 78/100
10 criteria × 10 points. Low pillars are visible on purpose — that's the point of a score you can audit. Methodology →
What it does well
- ▸Publishes 25 tokens/sec — the per-minute math is checkable: 1,500 tokens/min, promo ≈$0.014/min
- ▸≈$0.014–0.015/min across both Flash generations — third-cheapest verified per-minute in the paid table
- ▸Elo 1283 in the AA Speech Arena — competitive blind-vote quality
- ▸Commercial use from the API with no separate license tier
Where it costs you
- ▸3.8 Flash promo ($9/1M) doubles to $18/1M after 2026-12-31 — the post-promo ≈$0.027/min is the real budget number
- ▸No latency number published — throughput yes, latency no
- ▸No self-host path
Hidden traps — the stuff the pricing page doesn't say
Gemini 3.8 Flash TTS bills $9/1M audio tokens as a promo — it doubles to $18/1M after 2026-12-31. At the published 25 tokens/sec that is ≈$0.014/min today and ≈$0.027/min after: budget the doubled rate, because the promo is the number the pricing page shows you first.
Know the limits before you pay: read the traps above, then decide with eyes open.
Visit official siteAffiliate link where available — never affects the score.
- official Google AI pricing — audio token rates, promo terms (official, checked 2026-11-12) verified
- benchmark AA Speech Arena — Gemini 3.8 Flash TTS Elo 1283, $16.49/1M chars (snapshot 2026-11-12) verified
Data as of 2026-11-12. Prices and limits change monthly — report anything stale. How we verify: methodology.
Closest alternatives
MIT-licensed and CPU by design — the lightweight workhorse for the machine where there is no GPU at
Top-5 quality (Elo 1296) with the most honest meter in the business — the official ~750–800 credits/
Apache 2.0 weights with a confirmed 3-second rapid voice clone — the commercial-friendly cloning pic
Google Gemini TTS review
In-depth analysis · last updated 2026-11-12
The checkable token math
Google is the vendor that solves the token-opacity problem our hub complains about elsewhere: it publishes its audio rate — 25 tokens/sec — so the per-minute cost is arithmetic, not archaeology. 25 tokens/sec is 1,500 tokens/min; at the 3.8 Flash promo of $9/1M tokens, a finished minute costs ≈$0.014. Gemini 2.5 Flash TTS lands at ≈$0.015/min on the same math. Compare the row below this one: OpenAI bills audio tokens with no published rate at all.
The economics — and the expiry date
The catch has a date on it: the $9/1M promo doubles to $18/1M after 2026-12-31, which takes 3.8 Flash from ≈$0.014/min to ≈$0.027/min. The AA board lists Gemini 3.8 Flash TTS at $16.49/1M chars (Elo 1283) — consistent with the post-promo territory. At promo prices Google is the third-cheapest verified minute in the table (behind Simba ≈$0.005 and Fish Audio ≈$0.011); at post-promo prices it still undercuts ElevenLabs Creator credits (≈$0.14/min) by 5×.
Real limits
PAYG per audio token — no tiers, no published ceilings, no separate commercial license tier: the API terms are the terms. No latency figure is published (throughput is; latency isn't), and there is no self-host path. Quality sits at Elo 1283 — mid-board in the AA top-8, seven points behind Sonic 3.6, ten behind Simba 3.5.
How the pillars land
Price/Value 8.5: ≈$0.014/min verified at promo, with the post-promo ≈$0.027 docked in. Quotas 7.5: clean PAYG, no published ceilings, nothing exotic to game. Latency 8.5 is grounded in the published 25 tokens/sec throughput — the closest thing to a speed fact in the table after ElevenLabs' latency figure. Training use & voice cloning 6.5: no training-use or cloning policy facts in our snapshot — mid pillar pending audit. API & self-host 8: API only. Feature set 8.5: Elo 1283, two Flash generations to choose between. Track record 9: a published rate plus a dated promo expiry — predictable billing behavior, editorially scored. Ease 8: token-billed but the rate is published, so the math is one division. Pricing transparency 8.5: publishes what OpenAI won't — docked for leading with the promo number. Support & docs the neutral 5 — no support facts in our snapshot.
Verdict — GOOD, 78/100
The promo price makes it the thinking person's mid-table pick — checkable math, competitive Elo, no license-tier gymnastics. Two rules: budget the post-promo ≈$0.027/min, and re-check the promo before 2026-12-31. If the doubled rate sticks and your volume is high, the local table takes over; at low-to-mid volume, Google's arithmetic-honest billing is worth paying a little more for.