Google unveils Gemini 3.8 Live Avatar and prepares Gemini 4

Google has introduced Gemini 3.8 Live with Live Avatar, bringing real-time video avatars to AI conversations. The company also unveiled new text-to-speech models and confirmed that Gemini 4 is already in post-training, suggesting the next-generation model could arrive well before the end of 2026.

Gemini gets a real-time avatar

Live Avatar combines Gemini’s voice capabilities with low-latency video generation. The AI can listen, analyze visual information, and respond using a digital character with synchronized lip movements, facial expressions, and natural turn-taking.

Google is targeting the technology primarily at businesses, with potential applications ranging from customer support and virtual assistants to interactive training. Live Avatar is already available in Gemini Enterprise.

The system also supports background tool execution, allowing Gemini to perform tasks without interrupting the conversation. It can communicate in 97 languages and switch between languages during a single conversation.

Google says generated audio and video are protected with an invisible SynthID watermark to help identify AI-generated content.

New models can generate custom voices

Google has also introduced Gemini 3.8 Flash TTS and Flash-Lite TTS. The models allow users to control characteristics such as accent, emotion, pacing, pauses, and speaking style using natural-language instructions.

Google says its voice library includes more than 2,000 voices and supports over 100 languages and dialects. The models can also recreate a voice from a 30-second audio sample when the user has the necessary rights.

The new TTS models are being integrated into the Gemini API and Google AI Studio, with Flash TTS also coming to Gemini Notebook and Flash-Lite TTS to Google Vids.

Gemini 4 is already in post-training

Meanwhile, Google is preparing its next flagship AI model. Google DeepMind CEO Koray Kavukcuoglu confirmed that Gemini 4 has entered the post-training stage, which includes safety testing, behavioral tuning, and refinement.

Internal testing is already underway, including work related to Antigravity. Kavukcuoglu said Google hopes to release Gemini 4 significantly earlier than the end of 2026, although the company has not announced an exact launch date.

With Live Avatar, new voice-generation models, and Gemini 4 on the horizon, Google is pushing Gemini toward more natural, multimodal interactions while preparing its next major leap in AI capabilities.