Kubeez · Audio · Google Gemini 3.1 Flash TTS
Gemini 3.1 Flash TTS
Direct the read in plain words: tone, pace, emotion, accent
Google's Gemini 3.1 Flash TTS turns a script into a directed performance. Write a plain-language style prompt (calm, urgent, whispered, like a sports announcer) and drop inline tags like [laughing], [whispering], or [shouting] that the model actually performs instead of reading aloud. 30 voices, 70+ languages with auto-detect, billed from the same Kubeez wallet as ElevenLabs, music, and video. Section images below are Nano Banana 2 marketing stills; audio comes from the dialogue engine.
Showcase
DIRECTION, NOT JUST DICTATION
Tell the model how to read the line, then layer performed tags for laughs, whispers, and pace.
STYLE PROMPTS
Steer tone, pace, accent, and emotion in plain language ('warm and unhurried', 'breathless and urgent', 'like a 1940s newsreel') without hand-tuning sliders.
PERFORMED TAGS
Inline cues like [laughing], [whispering], [shouting], and [extremely fast] are acted out, not read literally, so reads land with real expression.
70+ LANGUAGES
Speak in 70+ languages with BCP-47 codes, or let auto-detect pick the language straight from your script.
Gemini TTS in the app and MCP
Use MCP tool "generate_dialogue" with the Google provider for automated Gemini voiceovers, with the same workspace and voices as Audio → Dialogue.
Where directed speech shines
Voices that act
Give a villain a sneer or a narrator a hush with a one-line style prompt, then revise the direction without re-recording.
Expressive in any language
Localise a hook into dozens of languages and keep the energy: auto-detect handles mixed-language scripts.
MCP tool "{{modelId}}"
Generate Gemini voiceovers from agents with the same generate_dialogue pipeline: select the Google provider and pass a style prompt.
Gemini 3.1 Flash TTS
Direct your voiceover with Gemini 3.1 Flash TTS
Open Audio → Dialogue, pick the Google provider, and write the read you want; credits stay in one wallet with ElevenLabs, music, and video.