Skip to main content

Gemini 3.8 Flash TTS API is the expressive speech option for acted dialogue and long narration: 70 voices, line-level direction and WAV delivery, with support for over 130 languages.

From $0.00032/s

Workbench

Input

Text to speech
Provide at least one of theseRequirement met
104/10000
Keep the Flash transcript within 10,000 characters and use <laugh>, <sigh> or <short pause> for local vocal events; separate character turns belong in dialogue. at least 1 characters · at most 10000 characters

About 9 sec of audio (rough estimate)

Dialogue0/20
Build Flash scenes with text for a solo reading or speakers and dialogue for up to two performers across 20 turns. Each turn supplies speaker, text and optional style, with a matching speaker name; text and each turn allow 10,000 characters. Speaker settings can override accent, style, pace and profile, while an omitted voice_id selects Kore. at least 1 items · at most 20 items
Speakers0/2
Build Flash scenes with text for a solo reading or speakers and dialogue for up to two performers across 20 turns. Each turn supplies speaker, text and optional style, with a matching speaker name; text and each turn allow 10,000 characters. Speaker settings can override accent, style, pace and profile, while an omitted voice_id selects Kore. Flash TTS exposes request-scoped cloning through a dialogue speaker’s reference_audio_url and consent_audio_url, without voice_id. Supply two WAV recordings of the same consenting adult, made with the same microphone in the same room, each no larger than 10 MB: 10–30 seconds of natural target-language speech for the reference and the required permission statement for consent. Synthetic recordings are ineligible, and no persistent voice ID is created; SpicyAPI has not yet validated this path with voice-owner recordings. at least 1 items · at most 2 items
Cast a Flash TTS narrator from 70 voices, including Fola, Neno, Charon and Algieba, with Kore used by default. Assign voice_id separately to dialogue speakers; cloning recordings replace that selection.
Style
WhisperEmpatheticVocal smileNewscasterPromo hypeDeadpan
Set style for the sustained acting of a Flash reading; an empty value keeps the voice’s delivery, and dialogue speakers may supply their own style.
Use accent to direct the English pronunciation of a Flash performance; leaving it out keeps the selected voice’s accent, and dialogue speakers inherit it only when they supply none.
0/1000
For expressive Flash performances, put sustained direction in instructions and reserve <laugh>, <sigh> or <short pause> for moments inside the transcript. The controls style, accent, pace and scene are combined into delivery notes; a dialogue line’s style takes precedence over its speaker’s style. A solo script stays a one-speaker performance, and each take needs listening for pronunciation or accidentally spoken directions. at most 1000 characters

Never read aloud. It only shapes how the script is performed.

Safety checkerOff

SpicyAPI does not run a safety checker on this model. Your prompt and media go to the model as written, and the model itself decides what comes back.

Checking your account before enabling submission and pricing.

Output

Ready for a new run
Example result
0:00 / –:––

Related models and next tasks

Missing the model you need?

70 voices, with direction for each line

Gemini 3.8 Flash TTS turns scripts into WAV speech with 70 prebuilt voices, including Fola and Neno, and delivery control for each line.

Open the playground
Prebuilt voices
70
Characters per request
10,000
Output
WAV
Languages
130+
Billing
Per sec
Platform filter
None
Prebuilt voices
Fola · Neno · Charon · AlgiebaOne of 70 prebuilt voices, including Fola, Neno, Charon and Algieba. Defaults to Kore. In dialogue, set voice_id on each speaker; omit it when using cloning recordings.
Delivery control
Delivery controlUse instructions for free-text direction; style, accent, pace and scene remain supported and are combined into each line’s delivery notes. In dialogue, a line’s free-text style overrides the speaker’s style. Single-voice requests use one speaker; no silent second speaker is added. Try native cues such as <laugh>, <sigh> and <short pause>. Delivery is generative: preview the audio to check pronunciation, cues and unwanted spoken instructions.
Dialogue and voices
Dialogue and voicesUse text for one voice, or speakers plus dialogue for a conversation. Each dialogue line has speaker, text and optional free-text style. Up to two speakers and 20 lines; text and each line accept up to 10,000 characters. A speaker’s name must match the line’s speaker. Each speaker may override accent, style, pace and profile. Omitted voice_id defaults to Kore.

What the Gemini 3.8 Flash TTS API does well

  • Audiobooks and long-form narration

    Chapters, bedtime stories and serialised fiction read by a steady narrator. Pick a voice once and keep it: the same voice sounds consistent from one chapter to the next.

  • Character and companion voices

    Give a chat character a voice of its own. A [whispering] cue and short, spoken-style sentences turn plain text into lines that sound meant for one listener.

  • Uncensored audio stories

    Unfiltered scripts are read as written. In our tests the model spoke every word of the monologues and narration we sent, and performed bracketed cues such as sighs and gasps rather than reading the tags.

  • Voice tracks for video

    Generate the line here, then hand the WAV to a talking-avatar or lip-sync model so a face performs it. The audio length sets the length of the clip downstream.

How to call the Gemini 3.8 Flash TTS API

  1. Dialogue and voices

    Use text for one voice, or speakers plus dialogue for a conversation. Each dialogue line has speaker, text and optional free-text style. Up to two speakers and 20 lines; text and each line accept up to 10,000 characters. A speaker’s name must match the line’s speaker. Each speaker may override accent, style, pace and profile. Omitted voice_id defaults to Kore.

  2. Delivery control

    Use instructions for free-text direction; style, accent, pace and scene remain supported and are combined into each line’s delivery notes. In dialogue, a line’s free-text style overrides the speaker’s style. Single-voice requests use one speaker; no silent second speaker is added. Try native cues such as <laugh>, <sigh> and <short pause>. Delivery is generative: preview the audio to check pronunciation, cues and unwanted spoken instructions.

  3. Billing

    Billed per second of audio returned. A hold for 660 seconds is placed while the task runs and the unused part is released when the audio is measured.

Gemini 3.8 Flash TTS worked examples

  • A whispered late-night voice note

    Use a short script to check the voice and delivery before generating a longer passage.

    {"text": "[whispers] Hey... I know it's late, but I couldn't stop thinking about you. Call me back when you get this.", "voice_id": "Enceladus"}

Gemini 3.8 Flash TTS API questions

Gemini 3.8 Flash TTS turns scripts into WAV speech with 70 prebuilt voices, including Fola and Neno, and delivery control for each line. Billed per second of audio returned. A hold for 660 seconds is placed while the task runs and the unused part is released when the audio is measured.

All Gemini models14

Text models9