Google’s AI videos get a big upgrade with Veo 3.1 (arstechnica.com)

Google has rolled out Veo 3.1, an upgrade to its text-to-video model that promises stronger prompt adherence, improved audio and overall realism, plus broader format support and a cheaper “Fast” variant for lower-cost generation. The model will be integrated across Google’s ecosystem—most notably in the Flow filmmaking tool—and is positioned to power vertical as well as landscape video formats, opening the door to YouTube Shorts and other short-form platforms. Google also leverages its massive YouTube corpus for training, which has already accelerated Veo’s progress since Veo 2 and 3. For the AI/ML community this matters on multiple fronts: better prompt fidelity reduces wasted compute and iteration, making video generation more efficient and practical for creators and researchers; improved audio modeling tightens the quality gap between generated and real footage; and native vertical support aligns the model with dominant social formats. The addition of a “Fast” mode signals attention to cost-performance tradeoffs, while broader distribution across Google products will increase both adoption and scrutiny—raising practical opportunities for creative tools and renewed concerns about misuse and deepfakes. The update also keeps Google competitive with rivals like OpenAI’s Sora in the rapidly evolving video-AI landscape.