Google Omni 1.1 Flash: Generates 10 seconds at a time, with continuous recording up to a maximum of 40 seconds

Perceive Beating AI News Flash: Google has released Gemini Omni 1.1 Flash, a new version of Gemini Omni Flash that was first unveiled in May and will have its API open for public testing by the end of June.

The maximum duration for a single video generation in the new version remains 10 seconds, but for the first time, video continuation has been introduced in the Omni API. After a video is generated, it can now be extended further. Each extension can add another 10 seconds, allowing the total length of the video after continuous extensions to reach 40 seconds. When extending, the model will reference the previous up to 10 seconds of footage to seamlessly connect characters, actions, and scenes.

Omni 1.1 has also added the ability to control the first and last frames. Users can provide an initial image and a final image, and the model will automatically generate the intermediate continuous video. For example, by specifying that a scene starts with a character far away and ends with the character approaching the camera, the model will fill in the entire movement process.

This update also introduces a more affordable 360p draft mode, with the system throughput reportedly increased by up to 60%, at a cost approximately one-third that of 720p. The API prices are as follows: 360p at $0.03 per second, 720p at $0.10, 1080p at $0.15, and 4K at $0.30. It is worth noting that 1080p and 4K are both upscaled, not native high-resolution outputs.