On September 24, 2026, Google DeepMind announced Gemini 3.8 Live with Live Avatar, a feature that adds a low-latency visual avatar to Gemini’s conversational AI, enabling real-time lip-synced expressions and natural turn-taking for more human-like interactions, especially in enterprise settings.
Short answer: On September 24, 2026, Google DeepMind announced Gemini 3.8 Live with Live Avatar, a feature that adds a low-latency visual avatar to Gemini’s conversational AI, enabling real-time lip-synced expressions and natural turn-taking for more human-like interactions, especially in enterprise settings.
On September 24, 2026, Google DeepMind announced the latest addition to its Gemini family, Gemini 3.8 Live with Live Avatar. The reveal came through a blog post authored by Shuo-yiin Chang, a research scientist, and CJ Zheng, a software engineer who spoke on behalf of the Gemini Audio Team. According to the announcement, the new feature builds on the momentum of the Gemini 3.8 Live launch that occurred the previous week.
The core idea behind Live Avatar is to give Gemini’s conversational abilities a visual dimension that appears in near real time. By tightly integrating the model’s live dialogue engine with a low-latency video stream, the system can generate a moving avatar that speaks, listens, and reacts simultaneously. This coupling is described as native, meaning the video generation is not a separate post-process but is woven directly into the interaction pipeline.
One of the highlighted benefits is the creation of a more natural and intuitive experience for both enterprise customers and end users. The avatar’s visual output is designed to match the spoken content with precise lip-syncing, while also displaying a range of natural facial expressions. The turn-taking behavior, how the avatar pauses, responds, and yields the floor, has been tuned to feel fluid, reducing the mechanical feel that can sometimes accompany text-only or audio-only agents.
Because the video is generated with minimal delay, the avatar appears to react almost instantly to user input. This near-real-time visual presence is intended to make interactions feel closer to a face-to-face conversation, even when the participant is interacting with an AI system. The announcement emphasized that the goal is to bridge the gap between purely verbal exchanges and the richer cues that humans rely on when communicating in person.
Enterprise applications were mentioned as a primary target. Companies that rely on virtual assistants for customer service, training, or internal collaboration could use the avatar to provide a more engaging interface. The visual cue can help convey tone and intent, potentially reducing misunderstandings that arise from text-only or voice-only interactions.
The technical approach described involves pairing near-real-time video generation with the existing speech output of Gemini 3.8 Live. The system creates a dynamic visual persona that is not a static image but a continuously updating representation that mirrors the flow of dialogue. This dynamic nature is what allows the avatar to exhibit expressions that shift with the conversation’s context, rather than staying fixed.
While the announcement did not disclose specific performance metrics such as frame rate or latency numbers, it stressed that the underlying infrastructure is designed to keep delays low enough for the visual feedback to feel immediate. The focus on low-latency streaming suggests that the team has optimized both the video generation pipeline and the delivery mechanism to maintain synchronization between audio and visual streams.
The introduction of Live Avatar also signals a broader trend within Google DeepMind’s research: moving beyond unimodal models toward systems that can handle multiple modalities in a tightly coupled fashion. By treating video as an integral part of the conversational loop rather than an add-on, the researchers aim to create agents that can perceive, produce, and respond to cues across sight and sound in a coordinated manner.
For developers and product builders who work with Gemini, the update means they now have access to a tool that can add a visual layer to their AI-driven applications without needing to stitch together separate components. The native coupling reduces integration complexity and could lower the barrier to creating experiences that feel more human-like.
In summary, the September 24, 2026 announcement introduced Gemini 3.8 Live with Live Avatar as a way to give Gemini’s conversational AI an immediate visual presence. By combining low-latency video generation with live speech, the feature delivers an avatar that lip-syncs accurately, shows natural expressions, and manages turn-taking smoothly. The aim is to make interactions with AI feel more like talking to a person, especially in enterprise settings where engagement and clarity matter. Builders interested in leveraging this capability should explore the updated Gemini APIs and consider how a dynamic visual persona could enhance their specific use cases.
Gemini 3.8 Live with Live Avatar is a feature that adds a near-real-time visual avatar to Gemini’s conversational AI, announced by Google DeepMind on September 24, 2026 in a blog post.
The announcement was made by Shuo-yin Chang, a research scientist, and CJ Zheng, a software engineer who spoke on behalf of the Gemini Audio Team, via a DeepMind blog post.
Live Avatar natively couples the model’s live dialogue engine with a low-latency video stream, generating a moving avatar that speaks, listens, and reacts simultaneously without a separate post-process.
It provides precise lip-syncing, natural facial expressions, and fluid turn-taking, making interactions feel more face-to-face; for enterprises it enhances customer service, training, and collaboration by conveying tone and intent and reducing misunderstandings.
The announcement did not disclose specific frame-rate or latency numbers, but stressed that the infrastructure is designed to keep delays low enough for the visual feedback to feel immediate, emphasizing low-latency streaming.
Source: Google DeepMind
On September 23, 2026, OpenAI launched MentalHealthBench, an expert-informed benchmark that evaluates how helpful and safe AI responses are in realistic mental-health conversations involving anxiety, depression, stress, and related topics.
California Governor Gavin Newsom signed seven bills on September 21 2026 that require data centers to disclose their electricity and water consumption. The legislation directs the Public Utilities Commission to set separate power rates, mandates monthly energy reporting starting in 2027, obliges operators to report water use when seeking permits or licenses, makes them fund needed water-infrastructure upgrades, and removes a prior exemption so each new project faces standard environmen
California Governor Gavin Newsom signed a package of bills on September 23, 2026 that requires data-center operators to disclose their electricity and water use, mandating monthly energy reporting and water-use reporting tied to permitting or licensing, while setting separate power rates and removing CEQA exemptions.
NO COMMENTS YET
Comments are open. Have a thought or a question? Share it below.