Google text-to-speech with natural single-speaker narration, selectable voices, and prompt-controlled style.
Audio prompts work best when they define mood, pacing, structure, and finish. The more clearly you describe the role of the sound, the cleaner the result tends to be.
Best results start with voice intent, pacing, and delivery style.
Gemini 3.1 Flash TTS Preview on Pixio is Google's low-latency text-to-speech model for natural single-speaker narration. Use it when you want expressive voiceover from text with a selectable prebuilt voice and prompt-controlled delivery.
Gemini 3.1 Flash TTS Preview on Pixio is Google's low-latency text-to-speech model for natural single-speaker narration. Use it when you want expressive voiceover from text with a selectable prebuilt voice and prompt-controlled delivery.
[whispers], [laughs], or [short pause].| Input | Notes |
|---|---|
| Text | The transcript to speak. Tags can be included inline. |
| Voice | One of Google's prebuilt Gemini TTS voices, such as Kore, Puck, or Sulafat. |
| Style Instructions | Optional performance direction for tone, pace, accent, or narrator style. |
Gemini 3.1 Flash TTS Preview costs 30 credits per 1,000 characters, rounded up per request.
| Need | Use |
|---|---|
| Google single-speaker TTS | Gemini 3.1 Flash TTS Preview |
| Voice cloning or custom voices | ElevenLabs TTS or MiniMax Voice Clone |
| Multi-speaker dialogue | ElevenLabs Text to Dialogue |
| Music generation | Pixio Music, Lyria, Songcraft, or Stable Audio |
Use production language, not just genre labels.
Tell the model how the energy should move over time.
For speech, define delivery style, tone, and pacing.
For music, define arrangement and emotional arc early.
A strong audio prompt describes role, pacing, tone, and finish so the output feels produced rather than generic.
Tell the model how the voice should land: tone, pacing, energy, and clarity.
Define how the piece should progress so the output feels intentional instead of flat or repetitive.
Use stronger prompts and cleaner references once the direction is already working.
Gemini 3.1 Flash TTS Preview is strongest when the brief is clear about function: what the sound should do, how it should move, and what it should feel like.
Use structure language early so the output lands closer to production-ready on the first passes.
For voice work, specify delivery and character. For music, specify arrangement and emotional progression.
Decide whether the output is carrying narrative, mood, rhythm, or all three.
Describe the build, energy, and transitions so the result has movement instead of flattening out.
Once the direction is right, refine and separate instead of regenerating blindly.
Pair voice generation with cloning when continuity across campaigns or characters matters.
Use generated music or speech as the finishing layer once the visual cut is already working.