Midjourney V1 Video

What it is

Midjourney released its first video model, V1, on 18 June 2025.^[1] It works as image-to-video: instead of starting from a written description alone, the user takes an existing picture — an image made in Midjourney or a start frame uploaded from elsewhere — and has the model animate it. A motion prompt describing how the shot should move is optional.^[1]

Beginning from a frame rather than from words is a real design choice. The first frame fixes the composition, the subject and the look of the shot, and the model's job is to carry that picture forward in time. Someone who already has an image they like can keep it and add movement, rather than describing it again in words and getting a different picture back.

How a job runs

Animation can be automatic, in which case the system writes the motion prompt itself, or manual, where the user supplies the prompt and chooses between low and high motion settings.^[1] Each job produces four 5-second videos.^[1] Any of those clips can be extended, about 4 seconds at a time, up to four times, for a total of about 21 seconds.^[1]

That structure — a batch of four short takes, each of which can be lengthened step by step — gives the user a set of variations to choose from before committing to a longer piece.

Cost and availability

A video job costs about eight times an image job, which Midjourney said is roughly the cost of one image per second of video.^[1] At launch the model was available on the web only, and a relax mode was being tested for Pro subscribers and above.^[1]

Where it sits in the company's plans

Midjourney said its goal is models capable of "real-time open-world simulations", combining image, video, 3D and real-time models.^[1] The company described the building blocks toward that goal as image models, video models, 3D models and real-time models.^[2] According to Midjourney, the next year would involve building those pieces individually, releasing them, and slowly combining them into a single unified system.^[2]

Within that plan, the video model was described as a technical stepping stone — something meant to be fun, easy, beautiful and affordable rather than a finished simulation.^[2] Read that way, V1 is less an end product than one of several parts being assembled: an image model that can be animated, released on its own, and later joined with 3D and real-time work.