Google has unveiled Veo 3.1, an enhanced AI video generation model that improves audio quality, realism, and prompt accuracy. The update expands availability across its ecosystem, including the Gemini app and Flow filmmaking tool, while introducing a cost-saving 'Fast' variant. This release aims to support vertical video formats popular on platforms like YouTube Shorts and TikTok.
Google's Veo 3.1 builds on the capabilities introduced with Veo earlier this year, which marked a significant leap in AI video quality following Veo 2. The new model emphasizes stronger prompt adherence, leading to more accurate video outputs and reduced computational waste. Audio features, a key aspect of the prior Veo 3 release, have been further refined for better integration.
Previously limited to 720p landscape videos, Veo 3.1 now generates both landscape and portrait 16:9 formats, aligning with the rise of vertical content online. This development supports Google's earlier commitment to integrate Veo tools into YouTube Shorts, which favor vertical videos similar to TikTok. As a result, AI-generated videos are expected to appear more frequently on such platforms.
The model rolls out across Google's AI tools: users can access Veo 3.1 and its 'Fast' variant via the Gemini app, while developers gain entry through Vertex AI and the Gemini API. The 'Fast' option helps manage costs on a per-token basis, potentially offering more generations for Gemini app users.
In the Flow filmmaking tool, Veo powers new features like Ingredients to Video, Frames to Video, and Extend, now compatible with generated audio. Users can upload images as references or endpoints while adding custom sound. Additional innovations include precision edits: replicating a video's look to add objects without altering the rest of the scene, available immediately in Flow and the API. Object removal follows a similar approach but launches in Flow soon.
This update positions Google competitively against rivals like OpenAI, which recently launched an iPhone app for its Sora video AI. The rollout begins today, urging caution when viewing online videos amid rising AI realism.