Gemini 3.8 Flash TTS

Gemini 3.8 Flash TTS is Google's text-to-speech model for expressive, natural-sounding speech from text. It follows natural-language delivery directions for tone, pace, and emotion, renders inline tags such as laughs, sighs, and pauses, and generates two-speaker dialogue in a single request. It detects the language of the text automatically across more than 130 languages, making it suitable for narration, voice agents, podcasts, audiobooks, and localized voiceover.

Complete technical specification for integration
Ready-to-use code snippets for common workflows
Step-by-step tutorials for advanced use cases
← All GuidesLanguages
How to localize speech with Gemini 3.8 Flash TTS, which detects the language from the text across more than 130 languages, and keep one voice in every market.
Introduction
Gemini 3.8 Flash TTS speaks more than 130 languages, and there is no language field to set: the model detects the language from the text. Write speech.text in the language you want spoken, and any of the 30 voices reads it.
नमस्ते! आपका ऑर्डर पैक हो चुका है और आज शाम सात बजे तक आपके घर पहुँच जाएगा। डिलीवरी के समय कृपया अपना फ़ोन पास रखें।
This guide covers the request, localizing one script with one voice, English terms inside another language and dialogue in another language.
The request
The request has no language parameter. The Hindi text is all that tells the model which language to speak.
import { createClient } from '@runware/sdk'
const client = await createClient({ apiKey: process.env.RUNWARE_API_KEY })
await client.connect()
const [result] = await client.run({
model: 'google:gemini@3.8-flash-tts',
speech: {
text: 'नमस्ते! आपका ऑर्डर पैक हो चुका है और आज शाम सात बजे तक आपके घर पहुँच जाएगा। डिलीवरी के समय कृपया अपना फ़ोन पास रखें।',
voice: 'Sulafat'
}
})import asyncio
import os
from runware import Runware
async def main():
async with Runware(api_key=os.environ["RUNWARE_API_KEY"]) as client:
results = await client.run({
"model": "google:gemini@3.8-flash-tts",
"speech": {
"text": "नमस्ते! आपका ऑर्डर पैक हो चुका है और आज शाम सात बजे तक आपके घर पहुँच जाएगा। डिलीवरी के समय कृपया अपना फ़ोन पास रखें।",
"voice": "Sulafat"
}
})
asyncio.run(main())curl https://api.runware.ai/v1 \
-H "Authorization: Bearer $RUNWARE_API_KEY" \
-H "Content-Type: application/json" \
-d '[
{
"taskType": "audioInference",
"taskUUID": "719272dd-f0ac-4b50-af79-14b8b25f4c94",
"model": "google:gemini@3.8-flash-tts",
"speech": {
"text": "नमस्ते! आपका ऑर्डर पैक हो चुका है और आज शाम सात बजे तक आपके घर पहुँच जाएगा। डिलीवरी के समय कृपया अपना फ़ोन पास रखें।",
"voice": "Sulafat"
}
}
]'runware run google:gemini@3.8-flash-tts \
speech.text="नमस्ते! आपका ऑर्डर पैक हो चुका है और आज शाम सात बजे तक आपके घर पहुँच जाएगा। डिलीवरी के समय कृपया अपना फ़ोन पास रखें।" \
speech.voice=Sulafat{
"taskType": "audioInference",
"taskUUID": "719272dd-f0ac-4b50-af79-14b8b25f4c94",
"model": "google:gemini@3.8-flash-tts",
"speech": {
"text": "नमस्ते! आपका ऑर्डर पैक हो चुका है और आज शाम सात बजे तक आपके घर पहुँच जाएगा। डिलीवरी के समय कृपया अपना फ़ोन पास रखें।",
"voice": "Sulafat"
}
}Response
[
{
"taskType": "audioInference",
"taskUUID": "719272dd-f0ac-4b50-af79-14b8b25f4c94",
"audioUUID": "360e54e5-4fa9-43c4-8772-a857cfd2098a",
"audioURL": "https://am.runware.ai/audio/os/a14d18/ws/2/ai/360e54e5-4fa9-43c4-8772-a857cfd2098a.mp3"
}
]Only speech.text changes from one language to the next, so a localization pipeline swaps the text and keeps the rest of the request as it is.
One voice for every market
A product video that ships in four markets can keep one narrator in all of them. Translate the script and send the same voice each time.
Meet Nimbus, the air purifier that learns your home. It runs quietly at night, speeds up when you start cooking, and clears the air in a large room in under ten minutes.
Te presentamos Nimbus, el purificador de aire que aprende cómo es tu casa. Funciona en silencio por la noche, acelera cuando empiezas a cocinar y limpia el aire de una habitación grande en menos de diez minutos.
Voici Nimbus, le purificateur d'air qui apprend à connaître votre maison. Il reste silencieux la nuit, accélère dès que vous commencez à cuisiner et purifie l'air d'une grande pièce en moins de dix minutes.
お部屋を学習する空気清浄機、ニンバスをご紹介します。夜は静かに運転し、料理を始めると自動でパワーアップ。広いお部屋の空気も、10分以内にきれいにします。
Store the voice name once with the product, and every new market is a translation and one more request.
English terms in another language
Product and interface names often stay in English inside a translated script. Write them as they appear on screen, and the model reads the sentence in the surrounding language with the names as written.
Para empezar, abre el Dashboard, entra en Settings y activa Smart Sync. A partir de ahí, tus pedidos y tu inventario se actualizan solos.
Dialogue in another language
A two-speaker request works the same way in any language. The aliases stay as declared in speech.voices, and the turns are written in the target language, such as this practice conversation for a language-learning app.
[Claire] Bonjour ! Je voudrais deux billets pour Lyon, s'il vous plaît. [Julien] Bien sûr. Aller simple ou aller-retour ? [Claire] Aller-retour, avec le retour dimanche soir. [Julien] Parfait. Ça fait quatre-vingt-six euros.
Tips
-
Write the text in the language you want spoken. Detection reads the text, so there is nothing else to set.
-
Keep one voice across markets. Every voice reads every language, so the brand sounds the same in each storefront.
-
Adapt the script for speech in each language. A translation written for a page reads as stiffly as English copy does, so apply Prompting to every version.
-
Keep inline tags in English. Inline tags covers tags inside text in other languages.