Google Launches Veo 3.1: Improved Model, Advanced Editing, and More Realistic Videos
Google continues to enhance its video generation model with the launch of Veo 3.1, an update that significantly improves audio quality, editing options, and the fidelity of rendered images.
This update follows Veo 3, introduced last May, and further solidifies Google’s position in the race for AI-generated video.
Veo 3.1: More Immersive Sound and Natural Videos
With Veo 3.1, Google introduces integrated audio generation into its AI clips. Each produced video can now include a soundtrack or contextual sound effects, making the output more vibrant and immersive.
The model also improves visual coherence and the realism of movements while better adhering to user-provided instructions—a major challenge for generative video models.
More Precise Editing Tools
Google emphasizes granular control over creation. Users can now:
- add an object to an existing scene and seamlessly integrate it into the clip’s style;
- soon remove objects directly within Flow, the video editor powered by Veo;
- extend a video from its last frames to smoothly prolong the sequence;
- or even use a reference image to animate a character or generate a scene from an initial and final frame.
All these features now support audio, enhancing the overall coherence between sound and image.
Deployment in Flow, Gemini, and Vertex
The Veo 3.1 model is being rolled out today across several products within the Google ecosystem:
- Flow, its intelligent video editor,
- the Gemini application,
- as well as the Gemini and Vertex APIs for developers.
Since the launch of Flow last May, Google claims that over 275 million videos have already been created on the platform.
Towards a New Era of AI Video Creation
With Veo 3.1, Google confirms its ambition to make multimodal video generation a central pillar of its AI strategy. By combining image, sound, and fine editing, the model positions itself as one of the most comprehensive on the market, competing with rivals like Runway Gen-3 Alpha or Pika 2.
