As Sora 2 After the launch,Google ‘s Veo 3 Discussion around video generation models has quieted down considerably. It seems that to keep this from continuing, Google quickly released the new Veo 3.1 earlier. Video generationThe model has been improved across the board, including better prompt comprehension, richer sound effects, stronger narrative control, and more realistic texture presentation. It is now available in Flow,Gemini and use on the Gemini API.
Google launches new Veo 3.1 video generation model — these are the new features
According to Google’s introduction: “Veo 3.1 is built on the foundation of Veo 3, with more precise prompt understanding and further improvements in video and audio quality.” Overall, the main upgrade is in the audio. As for the video side, my testing didn’t show much noticeable difference, but the understanding capability is indeed a bit better.
With the launch of Veo 3.1, making videos through Flow has also become easier—for example, you can now use:
- The “Asset-to-Video” feature lets you control characters, objects, and styles using multiple reference images, enabling Flow to generate final visuals that match your vision based on these assets.
- “Frame-to-Frame Video Generation” takes a starting and ending image, and Flow automatically creates a smooth transition video—perfect for artistic or epic scene transitions.
- “Extend Scene” allows videos to keep extending, even beyond one minute, and seamlessly connects to the last second of the previous video, making it perfect for shooting long takes or opening shots.
For the actual effect, you can refer to Google’s official video below:
In addition, Flow has also added two new editing features, including:
- “Add New Elements to Your Scene” helps you bring anything you can imagine into your shot—whether it’s real or a fantastical creature, Flow naturally handles shadows and lighting so new elements blend perfectly into the scene.
- “Seamlessly remove unwanted objects or characters” can be used to remove any element in the frame, and after removal, Flow will automatically reconstruct the background and surrounding environment.
Although Google’s official press release only mentions that Veo 3.1 is available to try on Flow and the Gemini API, I tested it and Gemini has already been updated to Veo 3.1 as well. If yours hasn’t yet, it should show up after waiting a bit.
If you see “New feature! Veo 3.1’s video generation is now even more powerful” at the bottom after opening Gemini, it means yours has also been updated to Veo 3.1:
The Flow section has already been fully upgraded to Veo 3.1, with the same two options: Veo 3.1 – Fast and Veo 3.1 – Quality. Also, you might notice that Veo 2 is nearing end of support — my guess is it might be replaced by Veo 3 at some point.
I also tested the differences in video generation between Veo 3 and Veo 3.1, and here are the results for your reference. Below is the prompt I used:
「一支高質感日系美食廣告短片。主角為一位 25 歲台灣優雅女性，白或米色短袖上衣、自然日系妝容與親切微笑。背景全程純色並保持乾淨：第一場景淡米黃、第二場景淺粉或薄荷綠。採柔和漫射光與微側光，色調明亮清新。鏡頭語言：每個場景一開始遠景，女士雙手端盤展示；展示時鏡頭平滑推近至食物特寫，景深淺，食物清晰、背景柔化；場景之間以快速白閃＋whoosh轉場。
場景 1：日式炸豬排飯，強調金黃酥脆麵衣、切面與醬汁光澤；女士微轉托盤角度配合高光。
場景 2：牛肉丼飯，強調油澤、洋蔥與溫泉蛋，可用筷子輕提牛肉纖維。
最後：回中近景，女士看鏡頭說中文『你喜歡吃哪一個？』。音樂為輕快清新日系 BGM，轉場有輕 whoosh。整體節奏俐落、色彩溫柔、畫面乾淨有食慾與幸福感
Veo 3 Fast video:
Veo 3.1 Fast Video:
Veo 3.1 Video:
Source: KOCPC Chinese