gemini-3.8-flash-tts

Text to Speech

Gemini 3.8 Flash TTS combines natural-language direction with expressive speech, letting you shape emotion, pace, and the delivery of each turn. On PoYo, choose from 30 preset voices and create narration or two-speaker conversations across 130 languages, bringing more vocal range to audio stories, lessons, and character dialogue.

Get APIKey

API Doc

Input

Text

Voice

Speakers

Speaker Id

Voice

Speaker Id

Voice

Turns

Turn 1

Speaker Id

Text

Style Instructions

Turn 2

Speaker Id

Text

Style Instructions

Enable Safety Checker

Billable characters: 265. Includes transcript and all style_instructions. Spaces, newlines and vocal-event tags count; speaker IDs do not.

Output

Example Output

This is sample data. Generate a new result to see real output.

Task ID:example-gemini-3-8-flash-tts-defaultCreated:2026-10-14T11:11:41

Status:finishedProgress:100%

Examples

Pricing details

Transparent pricing with no hidden fees. Pay as you go.

gemini-3.8-flash-tts

Text and style instructions

PoYo price

$0.045

9 credits

Official price

You save

Model	Spec	PoYo price	Official price	You save
gemini-3.8-flash-tts	Text and style instructions	$0.045 9 credits per 1000 characters	-	-

* Actual fees are based on the final output.

GEMINI 3.8 FLASH TTS

Use style instructions to control emotion and delivery for narration and dialogue

Gemini 3.8 Flash TTS API for Expressive Text to Speech

Make the delivery fit your script, from warm narration to expressive character dialogue. With Gemini 3.8 Flash TTS on PoYo, choose preset voices and use natural language to guide emotion, pacing, and per-turn delivery, creating more layered multilingual voice content.

Get API Key

Key Features of Gemini 3.8 Flash TTS API

Write the words, direct the delivery, and hear the difference. The clips below are Google official model demonstrations, not PoYo-generated examples.

Natural Emotion Beyond a Flat Read

The same service response can sound hesitant, apologetic, or reassuring. Gemini 3.8 Flash TTS controls delivery through natural-language instructions and allows momentary vocal events to be added inside the transcript. The official demonstration moves from a hesitant query to a reassuring reply, with pauses and sighs making the voice closer to everyday conversation.

Keep sustained emotion and pacing in style_instructions
Use inline events such as <sigh> and <short pause>
Choose from the 30 preset voices exposed on PoYo

Direct the Performance Line by Line

One delivery style rarely fits an entire scene. Keep the transcript separate from acting directions, then set global or per-turn style_instructions to guide emphasis, urgency, and pace. This official screenplay excerpt demonstrates how changing delivery and conversational reactions can build tension across successive lines.

Separate spoken text from delivery instructions
Set a shared style or direct individual turns
Preserve pauses and reactions where the script needs them

Two Distinct Voices in One Conversation

Build a dialogue without manually joining every spoken line. Configure exactly two speakers in speakers with their own preset voices, then submit ordered turns that reference their speaker_id values. The official table-read excerpt illustrates distinct vocal identities, natural turn-taking, and short listener responses in a performed conversation.

Configure two voices in speakers
Arrange the conversation with ordered turns
Add optional style_instructions to each turn

Who Can Benefit from Gemini 3.8 Flash TTS API?

Audiobook Producers

Give narration a clear emotional direction and distinct character voices. Gemini 3.8 Flash TTS lets audiobook teams separate spoken text from performance instructions, then reuse preset voices across their production workflow.

Podcast Teams

Turn a two-person script into a conversation with separate voices and ordered turns. Direct each host's delivery with Gemini 3.8 Flash TTS while keeping dialogue structure explicit in the request.

Game Dialogue Writers

Explore character interactions, reactions, and scene pacing with preset voices. Gemini 3.8 Flash TTS supports global and per-turn direction for testing dialogue before it moves into a game's audio pipeline.

Education and Training Teams

Create spoken lessons, role-play exercises, and narrated training materials. Use Gemini 3.8 Flash TTS to guide clarity and pacing while adapting written material for multilingual listening experiences.

Video and Voiceover Creators

Give explainers, product videos, and editorial scripts an appropriate tone. Gemini 3.8 Flash TTS keeps delivery instructions separate from the words, making voiceover revisions easier to manage.

Voice Application Developers

Add generated narration or scripted conversations to your product. Integrate Gemini 3.8 Flash TTS through PoYo's asynchronous task interface and retrieve audio when each generation finishes.

Gemini 3.8 Flash TTS vs Flash-Lite TTS

Choose the model for the workload. Flash focuses on creative performance; Flash-Lite focuses on efficient production at scale. The interface capabilities below reflect PoYo's current backend, while model positioning and language coverage follow Google's documentation.

Capability	Gemini 3.8 Flash TTS	Gemini 3.8 Flash-Lite TTS
Model focus	Voice fidelity, acting nuance, and regional expression	Efficient, high-volume speech production
Supported languages	130	101
Preset voices on PoYo	30	30
Single-speaker and two-speaker input	Supported	Supported
Global and per-turn style_instructions	Supported	Supported
PoYo integration	Asynchronous tasks with optional callbacks	Asynchronous tasks with optional callbacks
Typical workload	Audiobooks, narration, and expressive dialogue	Read-aloud features and bulk speech workflows

This compares two different models. No PoYo latency benchmark is implied. Google's voice design, voice replication, and extended voice library are not exposed by either of these PoYo endpoints.

How to Use Gemini 3.8 Flash TTS API on PoYo

Step 1: Register and create an API key.
Sign in to PoYo and create a key for your application.
Create API Key →

Step 2: Add credits.
Top up your balance and check the model pricing table. Transcript text and all style_instructions count toward usage.
Add Credits →

Step 3: Submit your script.
Send model: gemini-3.8-flash-tts with single-speaker text or structured speakers and turns to POST /api/generate/submit. Poll the returned task_id or supply callback_url to receive completion updates.
View API Reference →

Frequently Asked Questions about Gemini 3.8 Flash TTS API

What is Gemini 3.8 Flash TTS API?

Gemini 3.8 Flash TTS is Google's creative text-to-speech model for expressive narration and dialogue. PoYo provides asynchronous API access using 30 preset voices, separate style instructions, and structured two-speaker conversations. Submit your script with the model ID gemini-3.8-flash-tts and retrieve the generated audio when the task finishes.

How is usage calculated on PoYo?

Usage is proportional to billable characters, using the rate in the pricing table above. It includes transcript text, global style_instructions, and every turn's text and style_instructions. Spaces, newlines, emoji, and vocal-event tags count; speaker IDs do not. Requests are not rounded up to blocks of 1,000 characters.

Which voices can I use?

The current PoYo endpoint supports 30 preset voices, including Kore, Puck, Charon, and Zephyr. Single-speaker requests default to Kore. Each of the two speakers in dialogue mode selects one of those same presets. Custom voice creation, voice replication, and the extended voice library are not available through this endpoint.

How do I create a two-speaker conversation?

Provide exactly two speakers, each with a unique speaker_id and a supported voice. Then supply at least one item in turns with speaker_id and non-blank text. Add per-turn style_instructions when needed. Dialogue requests omit top-level text and voice, because those belong to single-speaker mode.

Where should I put emotion and acting instructions?

Keep sustained delivery requirements in global or per-turn style_instructions, while text only contains the content to be spoken. Momentary vocal events can be added in the transcript using tags such as <laugh>, <sigh>, or <short pause>, keeping delivery instructions and spoken content clearly separated.

How many languages are supported?

Google documents 130 languages for Gemini 3.8 Flash TTS, compared with 101 for Flash-Lite TTS. Language is detected from the transcript, so the current PoYo endpoint does not expose a language_code selector. Choose a supported preset and write the script in the language you need.

How do I receive the generated audio?

Submission returns a task_id. Query the existing status endpoint until the task is finished or failed, or provide an optional callback_url. Completed files are returned in data.files. The Playground lets you play supported audio and download the original file. This PoYo integration uses asynchronous tasks rather than audio streaming.

Are the feature clips generated on PoYo?

No. They are excerpts from Google's official Gemini 3.8 Flash TTS demonstrations. They illustrate the corresponding model capabilities and their sources are recorded in the asset manifest. PoYo-generated example outputs will be added after separate real generation and verification.

Why Use PoYo for Gemini 3.8 Flash TTS API Access

One Familiar API

Use PoYo's existing key and generation task interface to integrate speech alongside other supported models, with a consistent request envelope and task-status workflow.

Usage You Can Inspect

See the model's character-based rate and estimate usage before submission. Transcript text and style directions share one documented calculation, with no 1,000-character block rounding.

Test Before You Integrate

Explore single-speaker scripts and two-speaker dialogue in the Playground. Switch between form inputs and JSON to inspect the same request you will use in your application.

Fit Your Application Workflow

Track generation with task IDs and status polling, or provide a callback URL. Retrieve completed audio files when ready and keep generation separate from your application's request handling.