Gemini 3.1 Flash TTS¶

What it is¶

Gemini 3.1 Flash TTS is a next-generation text-to-speech model from Google, designed for low latency and high expressiveness. It is part of the Gemini 3.1 model family and supports native audio generation without requiring a Live WebSocket connection.

What problem it solves¶

It provides a way to generate high-quality, expressive AI speech with minimal latency, making it suitable for real-time applications, interactive AI assistants, and long-form content narration. It allows for fine-grained control over tone, pace, and emotion using steerable prompts and expressive tags.

Where it fits in the stack¶

AI & Knowledge / Generative Audio. It serves as the speech synthesis layer for multimodal AI applications.

Typical use cases¶

Interactive Assistants: Real-time voice interaction with LLM-based agents.
Content Creation: Generating high-fidelity voiceovers for videos, podcasts, or audiobooks.
Accessibility: Providing natural-sounding audio versions of text content with emotional depth.
Dynamic Narration: Games or apps that require context-aware speech.

Strengths¶

Low Latency: Optimized for fast response times and streaming.
Expressiveness: Supports emotional tags like [laughs], [sigh], and steerable prompts for style (e.g., "friendly and amused").
Multilingual: High-quality support for over 70 languages.
Integration: Uses the standard generate_content interface within the Google Gen AI SDK.

Limitations¶

Proprietary: Access is controlled by Google via their APIs (AI Studio / Vertex AI).
Rate Limits: Subject to Gemini API quota and usage-based pricing.
Preview Status: Some features and voices are in preview (as of 2026).

When to use it¶

When you need low-latency, high-quality speech synthesis within the Google ecosystem.
For interactive voice applications where emotional nuance and responsiveness are critical.

When not to use it¶

If your application requires a fully open-source or self-hosted TTS solution.
For offline use cases without internet access.

Getting started¶

Gemini 3.1 Flash TTS is accessible via the Google Gen AI SDK.

1. Installation¶

pip install google-genai

2. Hello World (Python)¶

from google import genai
from google.genai import types

client = genai.Client(api_key="YOUR_API_KEY")

config = types.GenerateContentConfig(
    response_modalities=["AUDIO"],
    speech_config={
        "voice_config": {
            "prebuilt_voice_config": {"voice_name": "Kore"}
        }
    }
)

response = client.models.generate_content(
    model="gemini-3.1-flash-tts",
    contents="Hello! Welcome to the future of text-to-speech.",
    config=config
)

# Save the generated audio
audio_bytes = response.candidates[0].content.parts[0].inline_data.data
with open("output.wav", "wb") as f:
    f.write(audio_bytes)

CLI examples¶

Programmatic access is primarily via REST. You can use curl to synthesize speech.

Basic Speech Synthesis¶

curl "https://generativelanguage.googleapis.com/v1beta/models/gemini-3.1-flash-tts:generateContent?key=$GOOGLE_API_KEY" \
    -H 'Content-Type: application/json' \
    -X POST \
    -d '{
      "contents": [{"parts": [{"text": "Synthesize this text into a warm, professional voice."}]}],
      "generationConfig": {
        "responseModalities": ["AUDIO"],
        "speechConfig": {
          "voiceConfig": {"prebuiltVoiceConfig": {"voice_name": "Callirrhoe"}}
        }
      }
    }'

Synthesis with Expressive Tags¶

curl "https://generativelanguage.googleapis.com/v1beta/models/gemini-3.1-flash-tts:generateContent?key=$GOOGLE_API_KEY" \
    -H 'Content-Type: application/json' \
    -X POST \
    -d '{
      "contents": [{"parts": [{"text": "[laughs] I did NOT expect that. [sigh] Can you believe it!"}]}],
      "generationConfig": {
        "responseModalities": ["AUDIO"]
      }
    }'

Synthesis with Steerable Prompt¶

curl "https://generativelanguage.googleapis.com/v1beta/models/gemini-3.1-flash-tts:generateContent?key=$GOOGLE_API_KEY" \
    -H 'Content-Type: application/json' \
    -X POST \
    -d '{
      "contents": [
        {"role": "system", "parts": [{"text": "You are a friendly and amused narrator."}]},
        {"role": "user", "parts": [{"text": "Tell me a quick joke about bananas."}]}
      ],
      "generationConfig": {
        "responseModalities": ["AUDIO"]
      }
    }'

API examples¶

Streaming Speech Generation¶

from google import genai
from google.genai import types

client = genai.Client()

config = types.GenerateContentConfig(response_modalities=["AUDIO"])

# Use generate_content_stream for low-latency playback
stream = client.models.generate_content_stream(
    model="gemini-3.1-flash-tts",
    contents="This is a long-form content that will be streamed in chunks.",
    config=config
)

for chunk in stream:
    if chunk.candidates[0].content.parts[0].inline_data:
        # Process audio chunk (e.g., feed to a player)
        audio_chunk = chunk.candidates[0].content.parts[0].inline_data.data
        print(f"Received audio chunk of size: {len(audio_chunk)}")

Sources / References¶

Contribution Metadata¶

Last reviewed: 2026-06-27
Confidence: high