June 19, 2012 - Popular AI image generation startup Midjourney today announced its entry into the AI video space with the launch of its first AI video generation model, V1, an image-to-video model that allows users to upload an image, or an image generated using Midjourney's other models, and V1 will generate four five-second videos based on that image. Similar to Midjourney's image model, V1 is currently only available through the Discord platform and will initially only be supported on the web.
Midjourney says it expects much more from its AI video models than just creating B-roll footage for Hollywood movies or generating ad videos for the advertising industry. In a blog post, Midjourney CEO David Holz said thatAI Video ModelIt's the next step in the company's journey toward its ultimate goal -- developing AI models that enable "real-time open-world simulations.
After the AI video model, Midjourney plans to further develop AI models for generating 3D renderings, as well as real-time AI models.
Notably, Midjourney's V1 model was released just a week before the company was sued by two of Hollywood's leading movie studios, Disney and Universal. The lawsuit claimed that Midjourney's AI image model generated images depicting the companies' copyrighted characters, such as The Simpsons' Homer Simpson and Star Wars' Darth Vader.
In terms of pricing, Midjourney says that it will charge eight times more for video generation than typical image generation, meaning that users will run out of their monthly generation quota faster when creating videos than when creating images. For now, the least expensive way to try out V1 is to subscribe to Midjourney's $10 per month (note: current exchange rate is about Rs. 71.9) base package. Users who subscribe to Midjourney's $60 per month Pro package and $120 per month Super package will enjoy unlimited video generation under the company's "Relax" model. Midjourney will re-evaluate its pricing strategy for its video model over the next month.
V1 offers users several custom settings to control the output of the video model. Users can choose between an automatic animation setting, which allows the image to move randomly, and a manual setting, which describes in text the specific animation they want to add to the video. In addition, users can adjust the amount of camera and subject motion by selecting either "Low Motion" or "High Motion" in the settings.
Although the videos generated by V1 are currently only five seconds long, users can choose to extend the video by four seconds up to four times, meaning that V1 videos can be up to 21 seconds long.
Similar to Midjourney's AI image model, early demos of the V1 video look to have a surreal aesthetic rather than extreme realism.Initial reaction to the V1 has been positive, but it's unclear how it compares to other leading AI video models that have been on the market for months or even years.