Gemini 3.8 Flash TTS

A sentence can welcome someone, explain a difficult idea, or leave a question hanging in the air. Start by deciding which of those jobs your recording needs to do. The voice gives you a starting character; the script and direction tell it how to approach the moment. Use a short voice description such as “warm and measured, with a curious finish” instead of stacking a long list of competing moods.

Gemini 3.8 Flash TTS supports the controls exposed in this studio: voice, accent, style, pace, a voice description, and scene direction. You can leave optional controls at their defaults. A plain sentence is often the most useful first test because it lets you hear the selected voice without several instructions pulling in different directions. Change one setting, generate another short take, and compare the same passage.

Read your text aloud before generating. Break long sentences where a person would naturally breathe, write unfamiliar abbreviations in a clearer form, and replace written shortcuts that sound awkward when spoken. Punctuation guides phrasing, while the supported inline tags provide specific cues for a pause, breath, laugh, or sigh. Put a cue where it belongs in the sentence, rather than collecting instructions at the end.

A short paragraph is enough to judge whether your opening sounds right. Listen for the pronunciation of names, the stress on important words, and the transition between sentences. If one line needs work, revise that line first. The generator accepts up to 5,000 characters per request; a shorter test leaves room to experiment before committing to a longer passage. Tags are part of the submitted script and contribute to its length.

For a conversation, switch to two speakers and choose a distinct voice for each participant. Assign every turn to speaker one or speaker two. A curious host and a measured guest are a useful starting pair: the contrast makes it easier to follow the conversation without exaggerating either performance. Scene direction can describe their shared situation, while each voice description describes only that speaker.

Keep an exchange focused on one topic. Start with a question, give the second speaker space to answer, and use a follow-up that responds to the answer. The studio submits the ordered turns as one dialogue and returns one audio result. It does not provide separate speaker tracks or an automatic video edit. For a reusable starting point, open the interview template and replace its lines with your own exchange.

The recording on this page is an English welcome generated for this site with Flash. It demonstrates one short take, not every voice or every possible delivery. The voice library contains neutral previews of all thirty available voices. Use those previews to narrow your choice, then test your own text in the studio: a voice that suits a friendly greeting may need different direction for a technical explanation.

A template is editable text and direction, not a pre-generated recording. Applying one does not use generation credits. You can change the voice, remove a tag, or rewrite a line before submitting. Playback and WAV download become available when your task completes, and completed recordings can also be found in your library.

Use Gemini 3.8 Flash TTS when you want to explore a performance with narration, character dialogue, or changes in tone. Flash-Lite is also available in the same interface and has a lower configured credit rate. Both models use the current voice and script controls. The best comparison is a short passage generated with the same voice and instructions, followed by a careful listen to the result.

The Flash-Lite page contains the detailed comparison and current credit examples. There is no need to rewrite a script to switch models, but switching does not regenerate an existing recording. Each submitted take is a new generation request. Check the estimate after changing the model or text, and keep your preferred WAV file before continuing with a new variation.