Midjourney has released Version 1 of its long-anticipated video model, introducing a new image-to-video workflow that allows users to animate still images with AI-generated motion.
The announcement was made by Midjourney founder David Holz, who outlined the broader vision behind the release, which is to assemble foundational components for real-time open-world AI simulations. These components include static image generation (already well-developed in Midjourney’s image models), motion (via video models), 3D spatial navigation, and real-time responsiveness. Over the next year, Midjourney plans to release each of these capabilities in modular form before gradually integrating them into a single, unified system.
The new video feature builds directly on Midjourney’s existing image generation workflows. Users can now animate any generated image using a new “Animate” button. This unlocks two modes: an automatic setting that creates movement based on inferred context and a manual mode that lets users write custom “motion prompts” to direct specific animation behavior. Two animation intensities, “low motion” for subtle, ambient movement and “high motion” for dynamic camera and subject motion, give users more control over output.
Midjourney also supports external content: users can upload images from outside the platform, designate them as a “start frame,” and animate them using the same motion prompt system. Videos can be extended in four-second increments up to a maximum of 20 seconds per sequence.
Midjourney, launched in 2022, has grown into one of the most widely-used generative image platforms. With a heavy focus on artistic coherence and user-friendly design, the tool quickly gained traction among both hobbyists and professionals. Unlike competitors that emphasize technical flexibility or multimodal integration, Midjourney has built its reputation on aesthetic output and community-driven iteration. The company’s expansion into video represents both a technical evolution and a necessary move to stay competitive in a space increasingly dominated by multimodal systems.
For the initial rollout, the video feature is available via Midjourney’s web interface only. Pricing reflects the increased computational load, with video jobs costing approximately eight times more than image jobs. Each video job generates four 5-second clips, effectively pricing video at about one image-generation unit per second, remarkably affordable compared to previous video AI models on the market. According to Midjourney, this represents a 25-fold cost reduction compared to existing commercial offerings. A “video relax” mode is also being tested for Pro-tier subscribers, offering a lower-priority, potentially cheaper processing option.
The company acknowledges that demand may outpace capacity during this early phase. Holz noted that pricing and availability could fluctuate over the coming month as usage patterns are analyzed and infrastructure is adjusted. Midjourney aims to strike a sustainable balance between accessibility and server load, especially given the high resource demands of real-time video synthesis.
Leave a Reply