Midjourney Unveils AI-Driven Video Generation Model V1 for Real-Time Simulations

The startup Midjourney has announced the launch of its latest large language model designed for video generation, named V1.

«As you may know, our focus for the past few years has been on [image generation]. However, what you might not be aware of is that we believe the ultimate purpose of this technology lies in models capable of real-time open-world simulations. […] From a technical perspective, this model is a step forward, but currently, we need to determine exactly what to provide you with,» the company’s press release states.

V1 is an AI-driven model that transforms images into videos. Users can upload a picture, and the AI will generate a set of four five-second video clips based on that image. If a particular clip is appealing, it can be extended.

The neural network comes with two different settings:

Additionally, the model features «low» and «high» modes. The low mode is better suited for scenarios with a static camera and a moving subject, while the high mode is optimized for more dynamic video content.

Currently, the model is in limited access for basic subscribers. Generating a video consumes eight times the internal «energy» of Midjourney compared to creating an image.

The startup’s team has promised to enhance performance and quality in the future, as well as introduce new features.

It’s worth noting that in June, Disney and Universal filed a lawsuit against the AI startup, alleging theft and copying of characters owned by the studios.