Gemini 3.8 Flash TTS
Advanced text-to-speech with natural language voice design and line-by-line performance direction
Context window
8192 tokens
Input / 1M tokens
$0.50
Output / 1M tokens
$9.00
Provider
Per 1M tokens through December 31, 2026: text input $0.50, audio output $9.00 (audio is 25 tokens per second, about $0.00225 per 10 seconds). From January 1, 2027: $1.00 / $18.00. Batch and Flex are half price. The cheaper Gemini 3.8 Flash-Lite TTS charges $6.00 per 1M audio output tokens. · Data verified 2026-11-08
Gemini 3.8 Flash TTS is a text-to-speech model built for deep creative direction and character design. It creates entirely new voices from scratch using natural language prompts and enables granular line-by-line control over acting cues, pacing, dialect shifts, and backchanneling across 130+ languages. Since November 4, 2026 it can also remix prebuilt or designed voices from natural-language instructions and save the result as a reusable voice ID.
How to access it
Available through Gemini API and Google AI Studio. Introductory pricing through December 31, 2026.
Strengths
- ✓Natural language voice design from scratch
- ✓Voice replication from 30-second audio samples
- ✓Fine-grained line-by-line performance control
- ✓130+ language support
- ✓Built-in SynthID watermarking for security
- ✓#1 on Hume AI Voice Design Benchmark
Best for developers who...
When to choose it (and when not to)
Reach for Gemini 3.8 Flash TTS when...
- →When voice quality and expressiveness are critical
- →When custom character design and acting direction are needed
- →For professional creative content production
Look elsewhere if...
- ✕When cost efficiency is the primary concern (Gemini 3.8 Flash-Lite TTS is cheaper at $6.00 vs $9.00 per 1M audio tokens)
- ✕For high-volume, commodity speech generation
- ✕When latency is time-critical
How to use it
- ›Describe voice characteristics in natural language for voice design
- ›Use performance direction annotations for line-by-line control
- ›Provide accent, tone, and emotional context in prompts
Quickstart
Pythonfrom google.generativeai import GenerativeModel
model = GenerativeModel('gemini-3.8-flash-tts')
response = model.generate_speech(text='Hello world', voice_config={'voice_description': 'young female, warm tone'})Requires google-generativeai library and API key
API model id: gemini-3.8-flash-tts
Benchmarks
| Benchmark | Score | Notes |
|---|---|---|
| Hume AI Voice Design Benchmark | #1 | Top ranking in voice design quality |
| Voice Arena | Near top | Near top rankings in Japanese, Hindi, and Mexican Spanish |
Source: DeepMind official announcement
Compare Gemini 3.8 Flash TTS
Compare Gemini 3.8 Flash TTS with any other model
Build a comparison →All model comparisons →