Google has launched Omni, a new family of generative AI models that can supposedly convert any input—photos, videos, text—into any other format. The first model in the lineup, Omni Flash, is now live in Google's Flow video generation and editing platform. According to Google, Omni outperforms the previous Veo model in several key areas: it accepts video uploads alongside text prompts to kickstart AI-generated creations, claims to embed more real-world knowledge into outputs, and maintains better character consistency throughout longer videos. The Verge's Allison Johnson put those claims to the test using her four-year-old's stuffed deer named Buddy. Last year, Johnson had attempted to recreate a Gemini advertisement featuring a plush animal on vacation, generating videos of Buddy at various destinations. This time, she brought Buddy back for a new round of AI-generated adventures to see if Omni could deliver improved results. The results were decidedly inconsistent. Johnson noted some outputs were "very good—much more consistent and true to" her original vision, while others fell short of the promises Google made about the model's capabilities. The mixed performance left her describing the experience as "baffling." The broader implication, however, was clear: AI video generation tools are becoming increasingly accessible, requiring minimal technical expertise to produce realistic-looking content. Johnson reflected that the line between harmless creative fun and what she called "full-on slop" may be harder to distinguish than ever, though she stopped short of declaring that distinction extinct entirely.