What is Gemini 3.5 Live Translate?
Gemini 3.5 Live Translate is a Gemini Live API model that does one job: interpreting. You stream speech in and name a target language, and it streams back translated speech, plus transcripts if you ask. It translates while the speaker is still talking and stays a few seconds behind, where turn-based voice models wait for each pause. It takes audio only and ignores instructions and tools, which is the main difference from Google's Gemini 3.8 Live.
It is the current version, in public preview on the Gemini API and Google Cloud, with no successor announced.
Where can I use Gemini 3.5 Live Translate?
On the Gemini API and Google Cloud for developers, in the Google Translate app for everyone, and in Google Meet for a small group of business customers.
- Gemini API and AI Studio
- Public preview for developers, with a free tier and a paid tier.
- Google Cloud
- Preview on Gemini Enterprise Agent Platform, formerly Vertex AI, in the global region.
- Google Translate app
- Live translate on Android and iOS with any headphones, plus a listening mode through the phone earpiece on Android.
- Google Meet
- Speech translation in 70+ languages, in private preview for select business Workspace customers.
Who is Gemini 3.5 Live Translate best for?
Gemini 3.5 Live Translate is best for developers adding live spoken translation to calls, meetings, lessons and broadcasts, and for travellers using Live translate in the Google Translate app.
Multilingual calls and meetings
Voice platforms such as LiveKit, Agora and Pipecat already connect to the Gemini Live API, so the audio streaming is ready-made.
Lessons, talks and broadcasts
Listeners hear the translation while the speaker keeps going, which suits one-way talks that rarely pause.
Rooms with mixed languages
Speakers can change language mid-session without anyone touching a setting.
Travel and face-to-face chats
The Live translate feature in the Google Translate app on Android and iOS uses it with any pair of headphones.
What's included with Gemini 3.5 Live Translate?
Continuous translation a few seconds behind the speaker, automatic detection of 70+ languages, a voice that follows the speaker, live transcripts, and settings you can lock for client apps.
- Continuous translation
- It speaks the translation while the speaker is still talking, a few seconds behind, instead of waiting for pauses.
- Automatic language detection
- It recognises the spoken language among 70+ on its own, so you only set the language to translate into.
- The speaker's voice
- Translated speech follows the speaker's intonation, pacing and pitch.
- Live transcripts
- Optional text of both the original and the translated speech, each tagged with its language code.
- Echo switch
- One setting decides whether speech already in the target language is repeated or left silent.
- Locked settings for client apps
- Browser and phone apps connect with short-lived tokens that can fix the target language, so your API key stays on your server.
- SynthID watermark
- Every clip of translated speech carries Google's inaudible SynthID watermark.
How does Gemini 3.5 Live Translate work?
Open a Live API session
Connect over WebSocket or a Google SDK with the model gemini-3.5-live-translate-preview and a target language code.
Stream the speaker's audio
Send raw 16-bit PCM audio at 16 kHz in chunks of about 100 ms.
Play the translation
Translated speech comes back as 24 kHz PCM audio, with transcripts of both sides if you turned them on.
How much does Gemini 3.5 Live Translate cost?
Gemini 3.5 Live Translate is free to try on the Gemini API's free tier, and the paid tier costs $3.50 per million audio input tokens and $21.00 per million audio output tokens, about $0.0368 a minute of translation.
- Free audio input and output
- Audio used to improve Google's products
- Try it in Google AI Studio
- About $0.0368 a minute of translated speech, input and output together
- About $0.0053 a minute in and $0.0315 a minute out
- Audio not used to improve Google's products
Google bills 25 audio tokens for every second of speech, both in and out. The model is in public preview on the Gemini Live API, and Batch, Flex and Priority tiers are not offered. The paid tier starts when you link a billing account and prepay at least $5 in credits, which expire after 12 months. Full pricing on ai.google.dev. We checked these prices on .
How do readers rate Gemini 3.5 Live Translate?
No one has rated Gemini 3.5 Live Translate yet. Be the first.
Sign in to rate. One rating per account.
Gemini 3.5 Live Translate pros and cons
It interprets continuously in 70+ languages for a few cents a minute, but it is a preview model that only translates, and Google warns that voices and language detection can slip.
Pros
- Translates while people talk rather than after each sentence, with automatic detection of 70+ languages.
- Free to try on the Gemini API, then about $0.0368 a minute with no plan or seat to buy.
- Works with voice platforms developers already use, including LiveKit, Agora, Pipecat and Fishjam, through the Gemini Live API.
- Two Hacker News users of Live translate in the Google Translate app, which runs on it, say it is hard to fault as a free service, and one calls it close to perfect for travel.
Cons
- Still a preview model with no open weights, and no Batch, Flex, caching or Provisioned Throughput options.
- Takes audio only and accepts no instructions, tools or custom vocabulary, so you cannot steer names or terminology.
- Google says voices can drift after long pauses or stick to one voice when several people talk fast, and one Hacker News user found the Translate app also picked up people talking in the background.
- Language detection struggles with heavy accents and close languages such as Spanish and Portuguese, and Google publishes no latency figure.
What are the best Gemini 3.5 Live Translate alternatives?
Gemini 3.8 Live for a voice agent that does more than translate, Pinch for a cheaper per-minute API, and DeepL Voice for ready-made meeting translation.
Gemini 3.8 LiveFreeChoose it when you need a voice agent that follows instructions, calls tools or searches the web, not only an interpreter; it answers turn by turn and costs $3 / $12 per million audio tokens.
PinchFreemiumChoose it when price per minute matters most: its live translation API costs $0.02 a minute against about $0.0368 here, though it covers 50+ languages rather than 70+.
DeepLFreeChoose it when you want translated meetings without writing code: DeepL Voice works in Teams, Zoom and Google Meet, keeps custom terminology, and starts at $13 per user a month for 300 minutes.
Questions about Gemini 3.5 Live Translate
What is Gemini 3.5 Live Translate?
Gemini 3.5 Live Translate is Google's speech-to-speech translation model for developers: it interprets 70+ languages while people talk, keeps each speaker's tone, and also powers Live translate in Google Translate.
Who is Gemini 3.5 Live Translate for?
Gemini 3.5 Live Translate is best for developers adding live spoken translation to calls, meetings, lessons and broadcasts, and for travellers using Live translate in the Google Translate app.
Is Gemini 3.5 Live Translate free?
Yes, developers can call Gemini 3.5 Live Translate free on the Gemini API's free tier within rate limits, where Google may use the audio to improve its products. Anyone can also use it through Live translate in the Google Translate app on Android and iOS.
How much does Gemini 3.5 Live Translate cost?
Gemini 3.5 Live Translate's paid plans are Paid Tier at $3.50 / $21.00 per million audio tokens in / out.
What are the best Gemini 3.5 Live Translate alternatives?
The closest alternatives on Recaply are Gemini 3.8 Live, Pinch and DeepL.
Where can I try Gemini 3.5 Live Translate?
On its official website, blog.google.
Can I use Gemini 3.5 Live Translate in Google Meet?
Only in a private preview. Select business Google Workspace customers get Meet speech translation powered by Gemini 3.5 Live Translate in 70+ languages, while Meet's standard speech translation still pairs English with six languages. Everyone else can use the model through the Google Translate app or the Gemini Live API.
What languages does Gemini 3.5 Live Translate support?
Gemini 3.5 Live Translate supports 70+ languages, from Afrikaans and Amharic to Vietnamese and Zulu, including Arabic, Hindi, Japanese, both Simplified and Traditional Chinese, and Brazilian and European Portuguese. It detects the spoken language itself; you only choose the language to translate into.
How is Gemini 3.5 Live Translate different from Gemini 3.8 Live?
Gemini 3.5 Live Translate is an interpreter: it translates speech continuously, takes only audio, and accepts no instructions or tools. Gemini 3.8 Live is a voice assistant that answers turn by turn, follows system instructions, calls tools and also reads text, images and video.
Does Gemini 3.5 Live Translate keep my voice?
Mostly. Gemini 3.5 Live Translate keeps the speaker's intonation, pacing and pitch, but Google warns the voice can shift after long pauses, start with the wrong gender, or stay on one voice when several people talk quickly.
Our verdict
Gemini 3.5 Live Translate is the simplest way to add simultaneous spoken translation to an app, as long as a preview model is acceptable.
It translates while people talk in 70+ languages, is free to try, and costs about $0.0368 a minute after that, but it only translates and Google gives no latency figure.
- Skip it for
- Voice agents that must also answer questions or call tools, and meetings where fixed terminology matters.
- Check first
- Test it with your speakers' accents and your noisiest rooms, since language detection and voice matching are its known weak spots.
TranslateGemma
Tiny Aya
TranslateImage
OpenL
Immersive Translate
TransGull