Google released the Gemini Omni 1.2 Flash video generation model on November 5, 2026. The model, referred to as Veo Omni Flash 1.2 in the SYNTX.AI service, supports video durations ranging from 3 to 15 seconds with one-second increments. It allows users to upload between 1 and 7 reference images to maintain consistency in characters, products, and styles.
In comparative tests, Omni Flash 1.2 scored 8.5 points in a text-to-video task, outperforming its predecessor, Omni Flash 1.1, which scored 7.0. The model also demonstrated improved performance in portrait generation, achieving a score of 86.5. However, it showed limitations in handling complex physics, such as liquid movement and object interaction, and occasionally struggled with maintaining details when processing multiple reference images.
The cost for generating an 8-second video is 21 tokens, matching the pricing of the previous version. For more complex tasks using the reference mode, the cost is 39.3 tokens per clip. While the model shows progress in consistency compared to earlier iterations, it still faces challenges with rapid motion, scene transitions, and maintaining specific details like tattoos or hair length across longer sequences.
The model is currently available through the SYNTX.AI platform. Despite improvements in text-to-video generation, the model remains behind top-tier competitors like Kling 4.0 and Seedance 2.5 in overall scene consistency benchmarks. Users are advised to verify output frames for advertising projects, as the model may occasionally deviate from provided reference designs.
