Google DeepMind has unveiled Gemini Omni 1.2 Flash, its next-generation video generation model. According to evaluations, the model ranks first in both the text-to-video and image-to-video categories and third in video editing, helping Google's Omni series sweep the top three spots on the text-to-video leaderboard. The model extends text-to-video duration to 15 seconds, supports native audio, can edit source videos of 3 to 15 seconds, supports clip extension up to 30 seconds, and enables retrospective editing via videoId, covering multiple resolutions and landscape/portrait aspect ratios. The model is now fully available on relevant platforms for direct use by all types of users, at a price 30% lower than the previous generation. Its pioneering draft-flow workflow—low-cost prototyping with on-demand upscaling—greatly reduces trial-and-error costs. New capabilities include multi-turn conversational fine-tuning and cinematic-level stable camera movement, strengthening features such as multimodal reference and multi-character audio-visual alignment, and precisely addressing past pain points of AI video such as the high cost of generating multiple candidates and the need to regenerate for modifications. Initial hands-on tests show outstanding performance in camera stability, content realism, and character consistency, marking that AI video development has entered a new stage of competing on price, delivery, and workflow adaptability.