Google has released the first real-world generated samples from its next-generation Veo 3.1 video model. This model has been integrated into its Vertex AI cloud platform and Google Vids video creation tool. Users can input a text prompt to generate 8-second, 720p video clips with accompanying soundtracks. Highlights of this update include a more engaging soundtrack and significantly improved visual fidelity and detail in the generated videos.
Leaked test videos, such as "Cyberpunk Hacker Robot" and "Volcano Eruption," demonstrate that Veo 3.1 understands prompts more accurately, capable of rendering intricate details like the flow of lava and realistic dinosaur formations. Compared to its predecessor, Veo 3, which occasionally suffered from repeated output and disproportionate object proportions, version 3.1 has significantly improved stability and accuracy.
This upgrade is seen by the industry as a significant step toward matching OpenAI Sora 2, with the quality of the videos it generates approaching professional production quality.
