Google rolled out Omni Flash this week, the first model in the company’s newly announced Omni family of generative AI tools. Integrated into Flow, Google’s AI‑powered video creation and editing platform, Omni Flash lets users upload a source video, add a text prompt, and receive a newly generated clip that blends the two inputs. The company says the model draws on broader real‑world knowledge and can maintain character continuity better than its predecessor, Veo.
To assess the claims, a tester used a plush deer named Buddy as a stand‑in subject. The results were a mixed bag. Some clips displayed tighter alignment with the prompt and smoother character motion than those produced by Veo five months earlier. In other cases, the deer abruptly changed orientation mid‑flight, antlers appeared out of nowhere, and objects such as a jar of honey morphed into a water bottle and back again. The model also struggled with maintaining consistent visual details when prompted to emphasize facial reactions.
Omni Flash also accepts text‑only prompts, allowing users to generate entire videos from scratch. When the tester asked the model to create a vacation montage for Buddy, the AI produced a playful sequence that, while entertaining, still featured noticeable “jump scares” and occasional visual artifacts. Editing existing footage with text commands proved more reliable than with Veo, yet the output still required multiple attempts to reach an acceptable level.
Cost remains a practical concern. Each video generation consumes between 15 and 40 credits, depending on length and the complexity of the input. Editing a clip costs a flat 40 credits. Google’s AI Pro plan, priced at $20 per month, supplies 1,000 credits, which the tester exhausted after roughly 20 clips, leaving just 145 credits for further experimentation. The pricing model could make extensive use of Omni Flash prohibitive for casual creators.
The model’s ability to blend AI‑generated elements into real‑world footage was also examined. Starting from a selfie video, the tester prompted Omni to place themselves eating spaghetti, sitting on an airplane, and standing before the Eiffel Tower. The resulting deep‑fake clips were convincing enough to fool a close family member, though subtle tell‑tale signs—such as an overly crisp fork‑clink sound and a background figure that repeated—betrayed the synthetic nature of the footage.
Overall, Omni Flash shows clear progress over Veo in terms of prompt adherence and visual consistency, yet it remains firmly within the “uncanny valley.” The technology makes it simple to craft short, playful videos, but the cost and occasional glitches keep it from delivering cinematic‑quality productions. As Google continues to refine the Omni family, creators can expect incremental improvements, but for now the tool is best suited for experimental or low‑stakes projects.
This article was written with the assistance of AI.
News Factory SEO helps you automate news content for your site.