Google Launches Real-Time Speech Translation via Gemini Live API

Google has added live speech translation to the Gemini Live API. With the new gemini-3.5-live-translate-preview model, developers can build real-time, speech-to-speech translations in more than 70 languages, including Dutch.

Google has implemented a major expansion for the Gemini Live API: real-time speech translation with extremely low latency, supported in more than 70 languages. With the new model gemini-3.5-live-translate-preview, developers can build applications that translate spoken language directly and fluently — without waiting for turn-taking in a conversation. This is an important milestone in the history of artificial intelligence in the field of multilingual communication.

What Is the Gemini Live API?

The Gemini Live API is Google's platform for real-time, multimodal interactions with the Gemini language model. The API is designed for applications where low latency is crucial, such as live conversations, interactive assistants, and — now also — direct speech translation. Unlike traditional translation services, which use text as an intermediate step, the Gemini Live API works entirely speech-to-speech: the user speaks, and the system returns translated speech almost instantly.

How Does Live Translation with Gemini Work?

The system continuously processes incoming audio in 100-millisecond chunks, without waiting for a speaker to finish. This ensures a smooth, uninterrupted translation experience that comes close to simultaneous interpretation by a human interpreter.

Technically, the API processes audio input as raw 16-bit PCM at 16kHz and returns the translated speech as raw 16-bit PCM at 24kHz — a higher output quality than the input. Languages are specified via international BCP-47 language codes (for example "nl" for Dutch or "es" for Spanish). Optionally, developers can also request transcripts of both the original and translated speech, which is useful for logging, accessibility, or subtitling.

More Than 70 Languages, Including Dutch

One of the most notable features is the broad language support: the Gemini Live API supports more than 70 languages, including Dutch, German, French, Spanish, Italian, Polish, and many more. For Dutch companies and developers, this is particularly relevant: Netherlands-specific applications — from customer service platforms to international conferencing tools — can now benefit from high-quality, real-time AI translation.

This aligns with the broader trend of AI applications that reduce the language barrier in business and personal communication. Google is positioning itself as a serious competitor to specialized translation services and real-time communication platforms.

Technical Capabilities for Developers

Google provides developer access through multiple programming languages and protocols. The Gemini Live API with live translation is available in:

• Python — via the official Google AI SDK
• JavaScript/TypeScript — for web and Node.js applications
• WebSockets — for direct, platform-independent integrations

A useful additional feature is echoTargetLanguage: this allows developers to specify whether speech already spoken in the target language is repeated or treated as silence. This makes the API flexible for both one-way translation and two-way conversations.

Limitations to Keep in Mind

Despite the impressive capabilities, the system also has some limitations. The API accepts only audio input — text input is not (yet) supported. Additionally, Google acknowledges that voice replication is not always consistent, and that automatic language detection may struggle with heavy accents or rapid language switches. For applications where precision is absolutely critical, such as medical or legal interpreting services, this deserves extra attention during implementation.

What Does This Mean for the Market?

The introduction of real-time AI speech translation via a developer platform like the Gemini Live API has far-reaching implications. Think of video calling software that automatically interprets, customer service bots that effortlessly switch between languages, or travel apps that communicate verbally in the destination country. Google makes this technology accessible to any developer with an API key, significantly lowering the barrier for innovative multilingual products.

Conclusion

With the Gemini Live API live translation feature, Google takes a major step toward seamless, AI-driven multilingual communication. The combination of low latency, broad language support — including Dutch — and accessible integration options makes this one of the most promising AI releases right now. Developers and businesses looking to take advantage of this technology would do well to start exploring this API today. Want to learn more about similar developments? Visit more AI news or deepen your knowledge through our knowledge base.

Google AI for Developers

---

Source: Google AI for Developers