Google Veo 3.1 Video Generation Model Released: New Audio Support, Object Addition, and More!

Today, Google announced the launch of a newVeo 3.1 video generation modelThe new version improves graphic realism and narrative control, as well as stronger cue word comprehension and synchronized audio/video generation.

Relying on the capabilities of Veo 3.1, Google has simultaneously upgraded Flow, its AI film and video creation platform. the biggest change in this update is in a number of core featuresAI-generated audio was introduced for the first timeThe creative experience of "Sound and Picture in One" is realized.

In Flow'sIngredients to Video (Ingredients to Video)In the feature, creators can upload multiple reference images and the system automatically generates a complete video with the specified character and scene styles.Frames to Video functionTransitional shots can be generated with natural articulation based on user uploaded footage.Video ExtendInstead, new footage can be extended from existing clips so that short clips can be connected to longer videos that last a minute or more.

The audio generation system automatically generates ambient sounds, action sounds and ambient soundtracks based on the content of the screen, making the work more immersive and tangible.

However, after generating a video, creators often have the need to further edit and modify the content of the screen. For this reason, Flow has added a number of new editing tools to give creators more precise control over the footage.

Creators can now add to existing videosInsert new element(e.g. characters, animals or props), the AI will automatically adjust the lighting and depth of field so that it blends naturally into the scene.Flow is also coming soon!One Click RemovalFunction, you can remove unwanted objects or characters, the system automatically rebuild the background, to achieve "no trace repair".

Veo 3.1 is now fully available to developers, businesses & creators through Flow, Gemini API, Vertex AI, and the Gemini app for a free trial. In Flow, Google is offering 100 AI points per month to buy free users, which can be used for the generation of approximately 5 videos.

Experience Links:https://labs.google/fx/zh/tools/flow

We experienced the generation effect of Veo 3.1. As you can see, Veo 3.1 offers two versions, namelyVeo 3.1-Fast with Veo 3.1-Quality.. In real-world testing, rapid generation takes about 30 seconds - 1 minute.

Veo 3.1 is able to accurately reproduce images involving complex light and shadow according to cue words, showing good light levels and natural atmosphere. The overall image is warm and soft, matching the true color temperature of early morning light. The water flow of the stream is simulated smoothly, with just the right amount of reflective detail. The voice-over is also in line with the content of the image and the cue words.

However, on the "hellishly difficult" gymnastics video generation, Veo 3.1 completely flopped, and the output was not in line with the human body's laws at all, with obvious distortions.

We also tried out the clip generation feature. After uploading coffee cups, tables, and portraits, Veo 3.1 can refer to the cue and the clip to generate the short advertisement that the user needs.Veo 3.1's rendering of the advertisement image and the advertisement lyrics conforms to the tonality defined in the cue, and also gives the advertisement lyrics a certain melody.

We uploaded a photo of OpenAI CEO Sam Altman, but the age and appearance of the people in the generated image differ from his. This may be related to Flow's protection mechanism for celebrity likeness rights: when uploading some images, the system prompts that the use of celebrity likenesses is not supported.

Conclusion: Technological advances are welcome, but still need to be applied with caution

From Sora 2 at the beginning of the month to Veo 3.1 today, video generation technology has progressed significantly in just a few weeks. However, the enhancements from Veo 3 to Veo 3.1 are not as significant compared to the leap from Veo 2 to Veo 3.

Overall, the release of Veo 3.1 provides a more powerful tool for AI video creation, but in practice, creators may still need to tweak the footage it generates to overcome current limitations and achieve higher-quality work.

Article source: Wisdom