Gemini 3.8 Flash-Lite TTS
Audio Available Full comparison ↗ google-deepmind/gemini-3.8-flash-lite-tts · by Google DeepMind
· audio-native-transformer
Pricing — 1 offering(s)
About this offering: Google's Gemini 3.8 Flash-Lite TTS, announced 2026-09-23 for high-volume speech such as voice agents and read-aloud features, in 101 languages. It uses the same API and voices as Gemini 3.8 Flash TTS and is billed per token: $0.50 input and $6 output per 1M, with Batch and Flex at half that, through December 31, 2026; Google's pricing page doubles all four rates from January 1, 2027 ($1 input and $12 output per 1M). Requests take up to 8,192 input tokens and return up to 16,384 output tokens.
Text input tokens
- $0.50 / 1M tokens (input) Current 2026-09-23 → 2026-12-31
- $1.00 / 1M tokens (input) Future 2027-01-01
Audio output tokens
- $6.00 / 1M tokens (output)
Current
2026-09-23 → 2026-12-31
Standard (paid tier) audio output rate; Google states it as equivalent to $0.0015 per 10s of audio (25 audio…
Standard (paid tier) audio output rate; Google states it as equivalent to $0.0015 per 10s of audio (25 audio tokens per second). Introductory rate through December 31, 2026; $12.00/1M ($0.003 per 10s) from January 1, 2027.
- $12.00 / 1M tokens (output) Future 2027-01-01
Batch
Batch text input tokens
- $0.25 / 1M tokens (input) Current 2026-09-23 → 2026-12-31
- $0.50 / 1M tokens (input) Future 2027-01-01
Batch audio output tokens
- $3.00 / 1M tokens (output) Current 2026-09-23 → 2026-12-31
- $6.00 / 1M tokens (output) Future 2027-01-01
Showing the active price and any recorded history. Full pricing history is available via the paid API — see API docs.
See how Gemini 3.8 Flash-Lite TTS fits into a cost-aware routing setup
See how →Capability profile
How Gemini 3.8 Flash-Lite TTS rates across core capability dimensions, with the task-level evidence behind each rating.
Google positions it for high throughput, low latency and cost efficiency rather than maximum fidelity. No independent naturalness score found as of 2026-09-27; not rated on vendor positioning alone.
30 prebuilt studio voices, plus an Extended Voice Library of hundreds more (GET /v1beta/voices) and Voice design personas generated from a text description (Google's speech-generation guide).
Voice replication from reference plus consent audio, stored persistently (up to 200 custom voices per project, 1-year retention) or as stateless client-held keys (Google's speech-generation guide).
101 languages per Google's model page, with automatic input-language detection.
Streaming supported (headerless 16-bit PCM at 24 kHz); Google publishes no time-to-first-audio figure.
Ratings are estimated — limited independent data is available for this model.
Operator guidance
Default Gemini TTS pick for cost-sensitive, high-volume or latency-sensitive speech ($6 vs $9 per 1M audio output tokens against Gemini 3.8 Flash TTS, introductory rates). Switching to Flash TTS is a one-parameter change, so step up only for narration, multi-speaker acting or the extra dialect coverage.
Use cases
- High-volume production speech generation
- Real-time voice-agent cascades (LLM → TTS)
- Read-aloud features and everyday single-speaker generation
Limitations
- No independent quality or latency benchmark found as of 2026-09-27 — capability ratings above rest on Google's own documentation
- Pricing is introductory through 2026-12-31 and doubles on 2027-01-01 (Google's pricing page)
- Token-billed: comparing against per-character TTS needs a tokens-per-character ratio Google does not publish (25 audio tokens per second of output is published)
- Inline text directions from earlier Gemini TTS prompts may now be spoken aloud — prompts need migrating to speech_metadata