内容来源：https://www.theverge.com/news/800371/google-veo-3-1-flow-audio

内容总结：

谷歌旗下AI视频制作工具Flow近日迎来重要更新，新版本通过提升光影处理能力和新增音频生成功能，使AI生成视频的逼真度达到新高度。

5月22日，谷歌宣布为Flow工具推出三项核心升级：用户现在可对AI视频中的光线效果与阴影投射进行精细化调整；新增"画面转视频"功能可将首尾两张图片自动生成连贯动态视频；"场景延展"功能则能在视频片段末尾续接最长1分钟的生成内容。这些功能均支持自动匹配生成背景音频。

此次升级同步对应Veo 3.1版本更新，该模型在根据图像提示生成视频的精准度方面获得提升。目前Veo 3.1已通过Gemini API向开发者开放付费预览，并在Gemini应用中启用。

值得注意的是，谷歌还预告即将推出视频元素智能擦除功能，该技术能重构被移除物体所在的背景场景，实现"无痕修图"效果。这一系列技术突破使得AI生成视频越来越难以被肉眼识别，在提升创作自由度的同时，也对内容真实性验证提出了新的挑战。

中文翻译：

谷歌公司正致力于提升其AI视频制作工具Flow生成视频的真实感，使其乍看之下更难被识别为人工智能产物。该公司周三宣布，用户现可为AI视频添加并调整阴影与光照效果。此次Flow工具的功能升级与Veo 3.1版本更新同步推出——谷歌表示新版本能更精准地根据输入图像素材生成视频内容。

借助Flow工具，用户可对AI生成视频的光照进行修正并添加阴影效果。

Flow用户还将通过多项新功能为视频生成配套音频。其中"素材转视频"功能支持用户上传三张参考图像并生成带音频的视频；"帧间动画"功能可基于起始与结束图像自动生成过渡动画及配乐；而"场景延展"功能则能选取视频末尾片段，自动生成最长60秒的延续内容并配备AI生成的音频。

Veo 3.1版本定价与Veo 3保持一致，目前作为"付费预览版"通过Gemini API向开发者开放，并在Gemini应用中启用。

谷歌透露，Flow工具即将推出"任意物体移除"功能——该技术将重构视频背景与场景，使被移除物体"如同根本不存在于原始画面中"。

英文来源：

Google is making videos created with the AI filmmaking tool Flow even more realistic — and harder to identify as AI-generated at first glance. The company announced Wednesday that users can add in and change the shadows and lighting of their AI videos. The expanded editing features in Flow are tied to the Veo 3.1 update, also announced on Wednesday, which Google says does a better job of making a video based on the images submitted as a prompt.
Google’s AI video generator is getting better editing and more audio
You can fix lighting and add shadows in AI-generated videos made with Flow.
You can fix lighting and add shadows in AI-generated videos made with Flow.
Flow users will also be able to generate videos with audio using several of the tool’s new features. Users can make a video with audio based on three reference images that the company calls “Ingredients to Video.” Another feature, called “Frames to Video,” creates a video that bridges a starting image with an ending image, with accompanying audio. The “Scene Extension” feature lets you take the final second of a clip and add on additional generated video up to a minute in length, also with generated audio.
Veo 3.1 costs the same as Veo 3, is available as part of a “paid preview” through Gemini API to developers, and is enabled in the Gemini app.
According to Google, Flow users will soon be able to remove “anything” from a video — the tool will restructure the background and scene to make “it look as though the object was never there.”