Midjourney has unveiled the initial version of its video generation model, making it accessible to users. Currently, the tool allows for the creation of short videos from images that have been uploaded or generated on the platform, with plans for additional features in the pipeline.
After users generate an image, a new “animate” button becomes available, enabling them to create a five-second video clip based on a text prompt. The platform also allows users to incorporate uploaded images as a “starting frame” for their videos. By default, the tool generates a basic prompt that “just makes things move,” but users can opt for a “manual” button to specify their desired animation style.
Users can extend their animations by four seconds, up to four times, allowing for a total video length of 21 seconds. Additionally, there are options for high and low motion settings that determine whether both the subject and camera move or just the subject alone.
The AI video generator is currently available on Midjourney’s website and through its Discord server. To access the service, users must subscribe, with prices starting at $10 per month, which includes 3.3 hours of “fast” GPU time sufficient for about 200 image generations. The startup has indicated that video generation will be priced approximately eight times higher than image generation, equating to roughly “one image worth of cost” for each second of video.
Midjourney is currently facing legal action from Disney and Universal, who raised concerns about the introduction of its video generator. The lawsuit claims that Midjourney operates as a “virtual vending machine,” producing unauthorized copies of copyrighted works from both companies. This controversy follows the announcement of the video generation model, which was first shared in January, with Disney and Universal alleging that the model’s training process likely infringes upon their copyrighted material.
In announcing the generator, Midjourney’s founder David Holz mentioned that this initial release serves as “a stepping stone” towards developing models capable of real-time open-world simulations. Other tech giants like Google, OpenAI, and Meta have also introduced their own AI video generators that leverage text prompts to create videos.
