Voice model · Google
Gemini Flash TTS is Google’s speech model; on Deepnia it runs Gemini 3.8 Flash TTS, generally available since September 22, 2026. You write up to 5,000 characters, pick one of Google’s prebuilt voices and place sounds such as a laugh, a sigh or a pause with English tags, such as <laugh>: the language is recognized from the text and you get a WAV file.
The voice generator opens with Gemini Flash TTS already selected. The exact cost shows before every creation; you sign up when you generate.
Google recommends 35 vocal tags for Gemini 3.8 Flash TTS, written in English between angle brackets: <laugh>, <sigh>, <gasp>, <short pause>, <long pause>… The model reads your script as written and plays each sound exactly where its tag sits. On Deepnia, the “Expression” panel adds laughs, sighs, pauses or whispers in one click, and you can type Google’s tags by hand.
Write in the language of your choice: Google announces more than 130 languages for Gemini 3.8 Flash TTS, and on Deepnia the language is recognized from your text, with nothing to set.
Choose from Google’s 30 prebuilt voices, each described by a style: Charon (informative), Kore (firm), Puck (upbeat), Sulafat (warm), Achernar (soft). Charon is the default voice.
Google presents Gemini 3.8 Flash TTS as its flagship speech model: studio-grade voice fidelity, expressive acting, authentic regional accents and strong stability over long texts. Enough to read a whole script with the same voice.
Pick a voice, write the text, generate: the language is recognized automatically and each generation delivers one WAV file (24 kHz, mono), ready for editing.
The “Try Gemini Flash TTS” button opens Deepnia’s voice generator with the model already selected.
Browse Google’s prebuilt voices (Charon by default), search by name or style and play their previews.
Write up to 5,000 characters, then add sounds from the “Expression” panel or type Google’s tags in English between angle brackets.
Check the cost shown, then click “Generate voice”: the language is detected automatically.
Copy a prompt, open the tool and adapt it to your project.
Once upon a time, there was a little fox who never wanted to sleep. <yawn> But that night, she found a hidden door in the old oak tree. <long pause> Behind it, a whole village glowed with fireflies.
<cheer> This weekend, it’s running week at your local sports store! <short pause> Shoes, watches and gear are waiting for you until Sunday night.
Breathe in deeply… <breath> then let your shoulders drop. <long pause> Let your breath find its own rhythm.
Press the green button, then wait three seconds before speaking. <short pause> The blue light confirms that recording has started.
Bienvenue dans Les Bricoleurs du week-end, le podcast des projets perso devenus sérieux. <laugh> Promis, cet épisode est plus court que le précédent.
So close! <sigh> The right answer was seven continents. <short pause> Next question: what is the longest river in the world?
Deepnia’s other voice and music models, to pick the right one for your project.
| Model | Inputs | Settings | Best for |
|---|---|---|---|
| Gemini Flash TTSThis model | Text | Expression tags | Expressive speech, with laughs, sighs and pauses exactly where you put them. |
| Open TTS | Text, Authorized personal voice | Language, Speed, Expression tags | Adjustable speed, a personal voice, 22 tags and SRT captions. |
| ElevenLabs | Text | Language, Stability, Expression tags | A 10-language menu and three stability levels. |
| MiniMax Speech 2.6 HD | Text, Authorized personal voice | Language, Speed | A cloned voice, with adjustable speed and 23 languages + Auto. |
| Deepnia Long-form | Text, PDF, Authorized personal voice | — | A whole document, text or PDF, read into a single MP3 file. |
Google announces more than 130 languages for Gemini 3.8 Flash TTS, French included. Write your text in the language you want: it is recognized from the text, and the tags are written in English.
Automatically: just write in the language you want and Gemini recognizes it from the text. Keep one language per script for consistent results.
30 Google prebuilt voices, each with a style and a preview: Charon (informative), Kore (firm), Puck (upbeat)… Charon is the default voice.
Add them from the “Expression” panel, or type Google’s tags in English between angle brackets, such as <laugh>, <sigh> or <short pause>, exactly where the sound should happen. Google recommends 35 of them.
Pick by style: Puck (upbeat) for an ad, Charon (informative) for a tutorial, Achernar (soft) for a bedtime story, Sulafat (warm) for a welcome message. Play the previews, then add tags where you want a sound.
Yes: Google designs Gemini 3.8 Flash TTS for studio-grade voice fidelity and strong stability over long texts. On Deepnia, you generate up to 5,000 characters at once with the same voice.
Gemini 3.8 Flash TTS, which Google made generally available on September 22, 2026: it reads your texts with Google’s 30 prebuilt voices and understands English vocal tags.
Up to 5,000 characters per generation, delivered as one WAV file (24 kHz, mono). For a longer text, generate it in several parts and join them in your editor.
The voice generator opens with Gemini Flash TTS already selected. The exact cost shows before every creation; you sign up when you generate.
