Google’s Veo 2 and Imagen 3 set new standards for AI. Should we be worried?

Google has unveiled Veo 2 and an upgraded Imagen 3, marking significant advancements in AI video and image generation. The announcement promises the tool to “achieve state-of-the-art results,” available through VideoFX, ImageFX, and a new Google Labs experiment, Whisk. Should we be worried?

Veo 2

Veo 2 pushes the boundaries of video generation by creating highly detailed, realistic videos in a variety of genres and styles. The model allows users to tailor their content by specifying camera angles, cinematic effects, and even mimic different lens types.

Unlike its predecessors, Veo 2 exhibits an improved understanding of real-world physics, human movement, and cinematographic techniques. This minimizes errors, such as the “hallucination” of extra fingers or unwanted details. Outputs can reach resolutions of up to 4K and extend to several minutes in length, making the tool suitable for both casual and professional projects.

Like other Google’s AI models, Veo 2 outputs include an invisible SynthID watermark. This helps identify them as AI-generated content, reducing the chances of misinformation and misattribution.

“YouTube creators are exploring the creative possibilities of video backgrounds for their YouTube Shorts, enterprise customers are enhancing creative workflows on Vertex AI and creatives are using VideoFX and ImageFX to tell their stories,” Google writes in a blog post. “Together with collaborators ranging from filmmakers to businesses, we’re continuing to develop and evolve these technologies.”

Imagen 3

Google also brings a new and improved Imagen 3 image generation model. It offers more vibrant, compositionally sound images that adhere closely to user prompts. Capable of generating diverse art styles—from photorealism to anime—the tool has been rated state-of-the-art in head-to-head comparisons.

The new Imagen 3 “renders richer details and textures,” Google notes. “In side-by-side comparisons of outputs by human raters against leading image generation models, Imagen 3 achieved state-of-the-art results.”

The model is now available to users in over 100 countries via ImageFX.

Whisk: A creative playground

Whisk is Google’s new experimental tool designed to enhance creativity by allowing users to remix existing images. It combines Imagen 3 with Gemini’s visual description capabilities to analyze and caption uploaded images. You can then manipulate these images, creating unique digital designs like stickers or enamel pins.

Whisk, initially available in the U.S., aims to expand accessibility to advanced AI tools while maintaining an interactive and fun user experience.

Implications for artists and creators

The introduction of Veo 2 and Imagen 3 brings both opportunities and concerns for the creative community. While these tools empower users to produce high-quality content effortlessly, they may have an impact on industries relying on traditional artistry. I’ve already read and heard many comments and thoughts from people using AI image generators. I recently read some Reddit posts about it and I was stunned to see how many people don’t hire illustrators, designers, or photographers any longer – they simply rely on AI.

Other than the displacement of human artists, there are also concerns over deepfakes, ethical misuse, and misinformation as generative AI tools become increasingly accurate and believable. Google’s commitment to responsible development and transparent labeling aims to mitigate such risks, but is it enough? I personally don’t think so, because we also need to rely on humans’ critical thinking and common sense, and I don’t think they’re as common these days when we’re bombarded with information.

While AI images already seem extremely believable, I think AI video generators won’t take our jobs. At least not for now.

[via Digital Camera World; lead image is AI-generated using Midjourney]

---