Google has launched Gemini 3.5 Live Translate, a new audio model offering low-latency, real-time speech translation across over 70 languages.

// 3h agoMODEL RELEASE

Google has launched Gemini 3.5 Live Translate, a new audio model offering low-latency, real-time speech translation across over 70 languages.

Google has released Gemini 3.5 Live Translate, a new audio model designed to process streaming speech in near real time, providing low-latency translation across more than 70 languages. The model is aimed at developers building real-time communication tools and is supported out of the box by ecosystem partners like Agora, LiveKit, Pipecat AI, and Software Mansion. Developers can test its capabilities within Google AI Studio's live playground.

// ANALYSIS

Native streaming audio-to-audio models are a major step forward, bypassing high-latency cascaded translation pipelines to enable true conversational speed.

–Native processing lowers latency and preserves speech nuances compared to traditional cascading speech-to-text-to-translation methods.
–Out-of-the-box partnerships with WebRTC platforms like Agora and LiveKit simplify production deployment.
–Translating across 70+ languages establishes strong coverage, though real-world accuracy will depend on audio quality and ambient noise.

// TAGS

googlegeminispeech-translationaudio-translationai-modellive-translatereal-time

DISCOVERED

3h ago

2026-06-09

PUBLISHED

3h ago

2026-06-09

RELEVANCE

8/ 10

AUTHOR

googleaidevs

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

UPDATE1h ago

CodeRabbit Slack agent generates developer visuals

CodeRabbit announced an update to its Agent for Slack, which is now capable of generating visual charts and graphs directly within the chat platform. The bot can pull data and create visualizations from services like GitHub traffic, BigQuery, and Datadog, providing teams with immediate visual context on their infrastructure and analytics without leaving Slack.

UPDATE1h ago

Factory integrates Claude Fable 5 into Droid

Factory has integrated Anthropic's new Claude Fable 5 model into its Droid autonomous software development agents to handle ambiguous, long-running engineering workflows. Early testing shows the integration improves Droid's autonomy, ability to recover from dead ends, and performance on complex debugging tasks.

NEWS1h ago

Claude Fable 5 guardrails block vulnerability audits

Users report that Anthropic's recently launched Claude Fable 5 model is flagging benign requests to check code for vulnerabilities. When strict classifiers for sensitive domains are triggered, requests route to the lower-capacity Claude Opus 4.8 fallback model, causing friction for developers.