Gemini 3.8 Flash TTS
Generate studio-grade voiceovers with Gemini 3.8 Flash TTS — shape emotion line by line, cast two speakers and cover 130 languages.
Support
Pro AI Tools
Explore elite tools including the Suno AI Music Generator.
MiniMax H3
MiniMax H3 AI Video Generator
Seedance 2.5
The Future of AI Video Is Here.
Seedance 2.0
The Future of AI Video Is Here.
FLUX 3 Video Generator
Kling 3.0
Next-Gen AI Video Generator
Grok Video Generator
Create Videos from Text or Images with AI
MiniMax H3 video generator
MiniMax H3 AI Video Generator
Seedance 2.0
The Future of AI Video Is Here.
Two Tiers, One Voice Engine: Inside the Gemini TTS Lineup
Launched on September 23, 2026, the Gemini TTS family pairs an expressive flagship model with a lean, high-volume sibling built for bulk narration.
- A Single Release, Two Very Different JobsThe flagship tier handles nuanced, character-driven reads, while its lighter sibling is tuned for fast, budget-friendly output at scale.
- You Direct the Delivery Instead of Choosing PresetsStyle notes attached to each turn, plus structured metadata and inline vocal tags, give you control over tempo, mood, accent and emphasis.
- Design a Voice or Replicate One With ConsentSketch a persona in plain words, or clone a real speaker using a clean reference clip together with a matching consent recording.
Prompting Rules That Keep the Audio Clean
Four habits that stop stray stage directions from leaking into the finished voice track.
What Gemini 3.8 Flash TTS Can Do: Feature Breakdown
From hands-on performance control to 130-language coverage, these are the strengths the flagship tier brings to production.
Hands-On Performance Control
Per-turn style cues plus inline laughs, sighs, coughs and pauses make it feel like coaching an actor rather than picking from a menu.
Voices Designed in Plain Language
Describe age, personality, accent, vocal texture and role in a prompt, with 2,000+ ready-made voices available through the Voices endpoint.
Cloning Behind a Consent Gate
Replication needs a clean reference clip and a matching consent recording from the same adult, plus SynthID watermarking and C2PA credentials.
Dialogue Scenes for Two Speakers
Write the exchange once and the model paces podcasts, teaching dialogues, product walkthroughs and game scenes — no manual line stitching.
Stable Long-Form Narration
Google documents consistent voice identity, timbre, loudness and room tone across multi-minute narration and extended conversations.
130 Languages and Regional Accents
The flagship tier reaches 130 languages while its lighter sibling stops at 101, with regional accents, minority dialects and IPA overrides.
Gemini 3.8 Flash TTS: Frequently Asked Questions
Quick answers on cost per minute, tier selection, benchmark standing and the safety rules around voice cloning.
What does it cost to generate a minute of audio?
Roughly 1.35 cents per audio minute, based on launch pricing of $0.50 for input and $9 for output per million tokens.
Which tier should I pick, Flash or Flash-Lite?
Reach for the flagship when acting nuance and long-form audio matter; choose Flash-Lite for bulk output and low latency.
How does it stack up against rival voice models?
Google cites 71.4 on Hume's Voice Design Benchmark, and Voice Arena places it second with 1,260 Elo.
What happened to the older Gemini 3.1 Flash TTS preview?
Flash-Lite TTS supersedes that preview and drops audio output pricing from $20 to $6 per million tokens.
Can I clone a voice, and what guardrails apply?
Yes, with a reference clip and a matching consent recording from the same adult speaker; SynthID and C2PA metadata travel with the audio.
Why does the model read my stage directions aloud?
Everything in the input is spoken verbatim, so move any lasting direction into the speech metadata fields instead.
Hear Gemini 3.8 Flash TTS Perform Your Own Script
Try both tiers in Google AI Studio or the Gemini API — moving from the flagship to Flash-Lite is nothing more than swapping a model identifier. Weigh batch against priority inference before you lock in a production budget.