What is Gemini 3.1 Flash TTS?
Gemini 3.1 Flash TTS is Google's text-to-speech model for the Gemini API. You send text and get audio back, from one voice or a two-person dialogue. Plain-language directions and inline tags such as [laughs] set style, pace and accent, it offers 30 prebuilt voices, and it speaks more than 70 languages. Every clip carries a SynthID watermark.
Google now lists it as a legacy preview. Gemini 3.8 Flash-Lite TTS is its direct replacement and costs less per audio token, and both 3.8 models add voice design and voice cloning. On Google Cloud, where the 3.8 models have not arrived yet, 3.1 is still the newest Gemini voice model.
Is Gemini 3.1 Flash TTS still available?
Yes, but Google now calls it a legacy preview and points new and production work to its Gemini 3.8 TTS models; no shutdown has been announced.
- Status
- Legacy preview on the Gemini API, with no shutdown announced.
- Direct replacement
- Gemini 3.8 Flash-Lite TTS, built for high-volume work, at $1 / $12 per million tokens.
- Flagship
- Gemini 3.8 Flash TTS, for the most expressive voices and long-form narration, at $1 / $18 per million tokens.
- Google Cloud
- Vertex AI and Cloud Text-to-Speech still offer 3.1 as their newest Gemini voice model.
Who is Gemini 3.1 Flash TTS best for?
Gemini 3.1 Flash TTS is best for teams already running it, or building on Google Cloud, who want directed, expressive speech in 70+ languages; new Gemini API projects should start on its 3.8 replacements.
Directed voice performances
Inline tags change delivery mid-sentence, and a scene description with director's notes sets tone, pace and accent for each speaker.
Two-person dialogue
One request can voice a conversation between two speakers, each with its own prebuilt voice, for podcasts or scripted scenes.
Apps in many languages
It speaks 70+ languages with control over style, pacing and accent for localized voice features.
Google Cloud projects
It runs through Vertex AI and Cloud Text-to-Speech, where Google's newer 3.8 voice models are not offered yet.
What's included with Gemini 3.1 Flash TTS?
Audio tags, plain-language direction, two-speaker dialogue, streaming, half-price batch jobs, AI Studio export and a SynthID watermark on every clip.
- Audio tags
- Bracketed tags in the text, such as [laughs] or [sigh], add reactions and change delivery at an exact point; Google lists more than 200.
- Plain-language direction
- A scene, a speaker profile and director's notes written in plain language steer tone, pace and accent before the transcript.
- Two-speaker dialogue
- Assign a prebuilt voice to each of two named speakers and the model voices the whole exchange in one request.
- Streaming output
- Audio can stream back in chunks while it is generated, so playback starts before the clip is finished.
- Batch API
- Half-price jobs for audio that can wait, with a 24-hour target turnaround.
- AI Studio export
- Voice settings tuned in Google AI Studio export as Gemini API code, so the same voice sounds the same across projects.
- SynthID watermark
- Every clip carries an inaudible watermark that marks it as AI-generated.
How much does Gemini 3.1 Flash TTS cost?
Gemini 3.1 Flash TTS is free to try on the Gemini API's free tier, and the paid tier costs $1.00 per million text input tokens and $20.00 per million audio output tokens, about $0.03 a minute of speech, or half that through the Batch API.
- Free text input and audio output
- Prompts used to improve Google's products
- No Batch jobs
- Text input, audio output; about $0.015 a minute of generated speech
- Asynchronous jobs with a 24-hour target turnaround
- Text not used to improve Google's products
- Text input, audio output
- About $0.03 a minute of generated speech
- Text not used to improve Google's products
Audio output counts 25 tokens per second, which is where the per-minute figures come from. Google now calls 3.1 Flash TTS a legacy preview and points new projects to Gemini 3.8 Flash TTS, listed at $1.00 per million text tokens in and $18.00 per million audio tokens out, or Gemini 3.8 Flash-Lite TTS at $1.00 and $12.00. The paid tier starts when you link a billing account and prepay at least $5 in credits, which expire after 12 months. Prices are in US dollars. Full pricing on ai.google.dev. We checked these prices on .
How do readers rate Gemini 3.1 Flash TTS?
No one has rated Gemini 3.1 Flash TTS yet. Be the first.
Sign in to rate. One rating per account.
Gemini 3.1 Flash TTS pros and cons
It is free to try and easy to direct, but it costs more than its 3.8 replacements, cannot clone voices, and users report drift on long clips.
Pros
- Free to use on the Gemini API's free tier, then about $0.03 a minute of speech on the paid tier, or half that on the Batch API.
- Plain-language directions and inline tags control style, pace and accent without extra tools, and one request can voice a two-person dialogue.
- Still offered on Vertex AI and Cloud Text-to-Speech, where Google's 3.8 voice models are not available yet.
Cons
- Several users report the voice drifting, running out of breath or turning tinny a few minutes into longer clips.
- It costs more than its own replacements: $20 per million audio tokens, against $18 for Gemini 3.8 Flash TTS and $12 for 3.8 Flash-Lite TTS.
- No voice cloning or custom voice design: you choose from 30 prebuilt voices, and one request handles at most two speakers.
- Google treats it as a legacy preview: no Flex or Priority tiers, and new and production work is pointed to the 3.8 models.
What are the best Gemini 3.1 Flash TTS alternatives?
Besides Google's own Gemini 3.8 TTS models: Eleven v3 for cloned voices and an app, Voxtral TTS for low-cost cloning and open weights, and Realtime TTS-2 for pay-as-you-go cloning and voice design.
Eleven v3FreeChoose it when you need cloned or designed voices and an app as well as an API; it costs $0.08 per 1,000 characters, about $0.08 a minute, in 70+ languages.
Voxtral TTSFreeChoose it when you want to clone a voice from a few seconds of audio for less, at $0.016 per 1,000 characters; it covers 9 languages and its open weights are for non-commercial use.
Realtime TTS-2FreemiumChoose it when you want voice cloning and voice design on pay-as-you-go billing; Inworld charges $25 per million characters with no subscription.
Questions about Gemini 3.1 Flash TTS
What is Gemini 3.1 Flash TTS?
Gemini 3.1 Flash TTS is Google's legacy text-to-speech model for developers, turning text into directed speech in 70+ languages with audio tags and two-speaker dialogue on the Gemini API.
Who is Gemini 3.1 Flash TTS for?
Gemini 3.1 Flash TTS is best for teams already running it, or building on Google Cloud, who want directed, expressive speech in 70+ languages; new Gemini API projects should start on its 3.8 replacements.
Is Gemini 3.1 Flash TTS free?
Yes, Gemini 3.1 Flash TTS is free on the Gemini API's free tier within rate limits, where Google may use your prompts to improve its products. Paid use costs $1.00 per million text tokens in and $20.00 per million audio tokens out, about $0.03 a minute.
How much does Gemini 3.1 Flash TTS cost?
Gemini 3.1 Flash TTS's paid plans are Batch at $0.50 / $10.00 per million input / output tokens and Standard at $1.00 / $20.00 per million input / output tokens.
What are the best Gemini 3.1 Flash TTS alternatives?
The closest alternatives on Recaply are Eleven v3, Voxtral TTS and Realtime TTS-2.
Where can I try Gemini 3.1 Flash TTS?
On its official website, blog.google.
Can Gemini 3.1 Flash TTS clone a voice?
No. It only uses Google's 30 prebuilt voices, and prompts change how a voice delivers lines, not whose voice it is. Voice design and voice cloning come with Gemini 3.8 Flash TTS and 3.8 Flash-Lite TTS, which ask for a consent recording before cloning a voice.
How long can a Gemini 3.1 Flash TTS clip be?
One request returns at most 16,384 audio tokens, which Google Cloud puts at about 655 seconds, or roughly 11 minutes; longer text is cut off. Split long scripts into several requests, which also helps with the voice drift some users report on long clips.
How do I move from Gemini 3.1 Flash TTS to Gemini 3.8 TTS?
Change the model ID, move style directions out of the transcript into speech_metadata, write inline tags in angle brackets such as <laugh>, and name the speaker on every turn. Gemini 3.8 returns WAV by default instead of raw PCM, so drop any code that adds a WAV header.
Our verdict
Gemini 3.1 Flash TTS still sounds good and is easy to direct, but new Gemini API projects should start on its 3.8 replacements.
Google lists it as a legacy preview, and Gemini 3.8 Flash-Lite TTS costs $12 per million audio tokens against its $20 while adding voice design and cloning.
- Skip it for
- Anyone who needs cloned voices, more than two speakers in one request, or clips longer than about 11 minutes.
- Check first
- Listen to your longest passages before you ship, since users report the voice drifting or turning tinny a few minutes in.
Gemini 3.1 Flash Live
Gemini 3.8 Flash TTS
Gemini 3.8 Flash-Lite TTS
Gemini 3.8 Live
Gemini 3.5 Transcribe
Chroma 1.0