Video is the new frontier for AI image generation models. OpenAI offers Sora, and Google has its Veo, two very impressive models that produce images on the edge of reality—some even go too far by exploring the desolate lands of the uncanny valley.
Midjourney slashes AI video prices
Midjourney couldn't possibly miss out. The still image specialist has therefore launched its first video model, V1, which generates 5-second animated clips (one job will deliver four 5-second videos, expandable up to 20 seconds). Unlike its rivals, V1 isn't necessarily aiming to produce photorealistic 4K videos; its goal is more modest: a simple, fast, and affordable tool for independent artists and those curious about AI.
This template is currently available in beta and web-only. Simply generate an image as usual, then click "Animate" to start an animation. Two modes are offered: automatic (the AI decides the movement) and manual (the user describes the movement precisely). V1 includes two animation styles. Low motion is used for calm or slow scenes (sometimes too slow, to the point of nothing moving at all). High motion is intended for dynamic scenes, with a greater risk of visual distortion.
Midjourney is banking on a very low cost: approximately one frame of credit per second of video, or 25 times cheaper than other similar AI services. This mainstream positioning contrasts with the competition's heavier and more expensive tools. The company is already planning major developments: 3D rendering, scene control, and immersive exploration are in the pipeline.
The company has a lot of ambition, the idea being to develop a system capable of generating open worlds in real time, nothing less. But for now, it's building the bricks one by one: the images and video are now available, there's still 3D to develop, then real-time AI.
Source: Midjourney
0 Comments