
据 The Verge 报道,谷歌最新的 Gemini 3.8 Live 更新允许用户与模型对话时,实时观看一个动画 AI 形象作出回应。新的 Live Avatar(实时数字人)功能可以在对话中实现口型同步,并展现不同的面部表情,但目前仅面向 Gemini Enterprise 企业客户开放。
谷歌表示,Live Avatar 可以在它支持的 97 种语言之间自由切换,且不会出现视频画质下降或视觉漂移的问题。谷歌分享的一段演示视频显示,该数字人分别用英语和日语讲话,口型动画与所说内容都能准确对上;它还能在讲话的同时在屏幕上调出相关信息。
Live Avatar 的发布距离谷歌推出 Gemini 3.8 Live 模型仅约一周,该模型据称可以近乎实时地处理视觉输入。谷歌将为客户提供一组预设形象库,同时允许企业自行创建专属数字人。谷歌还表示,Live Avatar 的输出内容将附带不可见的 SynthID 水印,并配备防止身份滥用的安全措施。
本文来源:https://www.theverge.com/tech/1000328/google-gemini-ai-live-avatar-face
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。
According to The Verge, Google's latest Gemini 3.8 Live update allows users to watch an animated AI avatar respond in real time while conversing with the model. The new Live Avatar feature enables lip-sync and various facial expressions during conversations, but is currently only available to Gemini Enterprise business customers.
Google says Live Avatar can switch freely among the 97 languages it supports without video quality degradation or visual drift. A demo video shared by Google shows the digital human speaking in English and Japanese, with lip-sync animation accurately matching the spoken content; it can also bring up relevant information on screen while speaking.
Live Avatar's release comes only about a week after Google launched the Gemini 3.8 Live model, which is said to process visual input in near real time. Google will provide customers with a set of preset avatar libraries while also allowing enterprises to create their own custom digital humans. Google also stated that Live Avatar output will include an invisible SynthID watermark and be equipped with security measures to prevent identity abuse.