Gemini 3.8 Flash TTS: write the voice, then write the script
Gemini 3.8 Flash TTS is Google's flagship voice model, and the first you direct rather than pick from. Describe the voice you want in plain language - a weary night-shift dispatcher, a bright teenage narrator - and it performs your script in that voice, across more than 130 languages, with the accent you asked for. Direct the delivery line by line, stage a two-hander between characters, and hold the same timbre steady across hours of audio. Available now in Picsart's AI Playground.
- Describe the voice you want
- Direct delivery line by line
Meet Gemini 3.8 Flash TTS, built for character, not just clarity
Most AI voice generators give you a list to choose from and a slider for speed. Gemini 3.8 Flash TTS is built the other way round: it is Google's top creative voice model, made for deep creative direction and character design, and it starts from a description rather than a dropdown. Google's own 30 original voices open out into a library of more than 2,000 production-ready ones, and you can write a new voice that exists nowhere in that library at all. It took first place on Hume AI's Voice Design Benchmark, which is the exact thing this model is for.
Design a voice in a sentence, in over 130 languages
Write what you want to hear and the model builds it: age, texture, energy, and the accent it should carry. Ask for Mexican Spanish rather than generic Spanish, Quebec French rather than Parisian, Scots or Southern US rather than a flattened neutral. It runs across more than 130 languages and works out which one you have written in on its own, so a script does not need to be tagged or declared - with its strongest results in Japanese, Brazilian Portuguese, Vietnamese, Modern Standard Arabic, Mexican Spanish and Hindi. The same description gives you the same character every time, which is what makes a voice reusable across a series rather than a one-off.
- Shape regional accents
Give notes to an AI voice the way you would to an actor
A flat read is the thing that gives synthetic speech away. Here you direct each line separately: play this one tired, land that one as a punchline, take the next one down to a near-whisper. Drop in a laugh, a sigh, a gasp or a caught breath exactly where the performance needs it. Add the sounds real conversation is made of - the murmured agreement, the small noise someone makes while the other person is still talking - and stage a genuine two-hander where each character holds their own voice through the exchange. It is closer to directing a read than generating one, which is why the output stops sounding like text to speech.
A voice that does not drift over a long recording
Long-form is where most voice models come apart: the timbre wanders, the room changes under the speaker, chapter nine does not match chapter one. Gemini 3.8 Flash TTS holds voice quality, pacing and character timbre steady across hours, keeping the same volume and the same acoustic space from the first line to the last. Audio arrives as a standard WAV, ready to drop into an edit. Working at volume instead, on a dubbing run or a voice agent? Gemini 3.8 Flash-Lite TTS is the same family tuned for throughput. Inside Picsart both sit alongside 150+ models, so you can voice a script here and switch models for the next one.
- Keep one voice across long recordings
What you can make with Gemini 3.8 Flash TTS
Audiobooks and long narration
Record hours in one consistent voice without the timbre sliding between sessions, and give the narrator direction chapter by chapter.
Gemini 3.8 Flash TTS FAQ
It is Google's flagship text-to-speech model, released in September 2026 and built for creative direction and character design. Rather than picking from a fixed list of voices, you describe the voice you want and it performs your script in it, across more than 130 languages, with direction applied line by line.
Generate AI visuals with GPT Image 2.5
Pro
Most popularAI tools for everyday creative work.
Ultra
Most powerfulHeavy AI usage for creators & teams.
Enterprise
Custom AI solutions for large organizations.
- Volume discounts on credit rate
- On-demand top-ups
Generate AI visuals with GPT Image 2.5
Pro
Most popularAI tools for everyday creative work.
Ultra
Most powerfulHeavy AI usage for creators & teams.
Enterprise
Custom AI solutions for large organizations.
- Volume discounts on credit rate
- On-demand top-ups