Midjourney has announced the launch of its first proprietary video model, marking a significant milestone in the development of artificial intelligence (AI) systems that can realistically model entire 3D worlds in real-time. The feature, dubbed “Image to Video,” allows users to convert any image created with Midjourney into a short animated clip. This capability is facilitated by the introduction of the new “Animate” button in the web interface of the service, as reported by TechCrunch.
Users can choose between an automatic mode, where the system determines the movement, and a manual mode, allowing them to describe how they want the animation to unfold. Additionally, images created outside of Midjourney can also be animated by dragging them into the prompt panel and selecting them as the “initial frame.” The desired movement can then be articulated via a text prompt.
Currently, the animation feature is only available through Midjourney’s web interface. Each generated video comes with a cost, approximately eight times higher than that of still images, returning four five-second clips. However, Midjourney claims that this pricing is roughly 25 times cheaper than that of its competitors. The startup has introduced its first video model, V1, which converts images into short videos. Users can upload their own images or select created ones within Midjourney, yielding four animation alternatives, each five seconds long. This new function operates through Discord and supports both automatic movement generation and manual adjustments using text prompts. The launch coincides with ongoing legal battles with Disney and Universal, which have accused Midjourney of copyright infringements.
The new feature, which enhances the existing image generation process, now allows for a seamless transition from still images to videos. After creating an image from a text prompt or uploading one, users simply need to press the “Animate” button to create videos. Two movement modes are available: the “low” mode is best for calm or minimalist effects, while the “high motion” mode aims for more dynamic animations, although it may lead to visual errors. While users can extend each video clip’s duration by four seconds, the maximum length remains capped at 20 seconds. Future updates are expected to enhance duration and functionality.
The cost of creating videos is eight times higher than generating regular images, meaning subscribers will use their monthly generation limits for videos much faster. The most accessible way to try the new feature is by subscribing to Midjourney’s basic plan at $10 per month, with higher-tier plans allowing for unlimited video creation in a slower “Relax” mode. Midjourney has also pledged to reassess the pricing of its video models within the next month.
The first examples of video V1 appeared more fantastical than realistic, and initial reactions have been generally positive. However, it remains unclear how well the new model will compete with other established AI video services currently on the market, such as Sora from OpenAI, Gen 4 from Runway, Firefly from Adobe, and Veo 3 from Google. Whereas most companies focus on developing video models for commercial purposes, Midjourney is positioning itself uniquely by offering creative AI solutions for creative professionals.
CEO of Midjourney, David Holz, explained in the company’s blog that the development of video models is just a stepping stone toward a more ambitious goal: creating an AI capable of “realistic modeling of open worlds in real time.” Following the video models, Midjourney plans to concentrate on developing AI for generating 3D graphics and establishing algorithms that function in real-time.