Google has launched Gemini Omni, a new AI model designed to generate “anything from any input,” with a focus on video creation. The first implementation, Gemini Omni Flash, is being rolled out to the Gemini app, Google Flow, and YouTube Shorts. Google announced that this model represents “the next step” beyond its previous models, Nano Banana and Veo 3.1.
Gemini Omni allows users to combine images, audio, video, and text to produce high-quality videos informed by real-world knowledge. Users can edit these videos using natural conversation, where each subsequent instruction builds on the previous ones, ensuring consistency in characters and other elements.
Unlike Veo 3.1, which could only generate videos from prompts and images, Gemini Omni accepts a broader array of inputs. Users can simply start with a live video and request changes, such as altering the action or adding characters. Google explained that users can transform their footage into something “you never could have filmed yourself.”
The model understands physical forces, including gravity and kinetic energy, which contribute to more realistic video scenes. Additionally, it integrates Gemini’s knowledge of history, science, and cultural references to enhance storytelling. The app can generate visual explainers from short prompts, although it will initially support only voice references for audio output.
Gemini Omni also lets users create a digital avatar using their voice. Google stated that it has established policies to protect users from potential harm associated with the use of its AI tools. Editing features regarding audio and speech are under testing for a responsible rollout.
Videos generated through this model will include Google’s SynthID digital watermark, aimed at verifying their origin with Gemini Omni. Concerns regarding the “uncanny valley” effect seen in previous models like Veo 3.1 have been noted, as many users found that quality undesirable.
Gemini Omni Flash is being made available to Google AI Plus, Pro, and Ultra subscribers globally, with the rollout to users on YouTube Shorts and the YouTube Create App commencing this week.
