# What is Midjourney's V1 video model, and how does it turn a picture into a clip?

Date: 2025-06-19 · Author: Robert Scales · https://guidetoartificialintelligence.com/en/learn/what-is-midjourney-v1-video-model/

**Answer:** Midjourney's V1 video model, released on June 18, 2025, takes a still image, either one you made in Midjourney or one you upload, and animates it into four short five-second clips that you can extend to about 21 seconds. It is web-only, works on every paid plan starting at $10 a month, and costs roughly eight times as much as an image job.

Yesterday, June 18, 2025, Midjourney released Version 1 of its video model to its whole community. Until now, the company has been known for still pictures: you type a description and it paints four images. With V1, there is a new button under your images called **Animate**, and pressing it turns a picture into short moving clips.

The idea is simple to state. You do not describe a whole film. You start from one frame you already like, and the model imagines what happens next: the wind moving through grass, a character turning their head, the camera drifting forward.

## What Midjourney released, and when

The announcement came from David Holz, Midjourney's founder and CEO, in a short post on the company's updates site. The release is open to everyone with a subscription, and for now it runs only on the Midjourney website.

Here is what one video request, called a **job**, gives you:

| Item | What you get at launch |
|---|---|
| Starting point | An image made in Midjourney, or your own picture uploaded as a "start frame" |
| Output per job | Four clips, each five seconds long |
| Extensions | Add about four seconds at a time, up to four times (about 21 seconds in total) |
| Cost | About eight times the cost of an image job |
| Cheapest access | The Basic plan at $10 a month |
| Unlimited slow mode | Being tested for Pro ($60 a month) and Mega ($120 a month) subscribers |

Holz says Midjourney finds it hard to predict what video will cost to run, so prices may change over the next month as the company watches how people use it.

## How image-to-video works

A video is a series of still pictures, called frames, shown quickly one after another. An **image-to-video** model receives the first frame and has to invent the ones that follow, so that the scene moves in a believable way while still looking like your original picture. The model has learned, from huge amounts of training material, how things usually move: how hair falls, how waves roll, how a camera pans.

You guide that guess in two ways.

**Automatic or manual.** In automatic mode, Midjourney writes a **motion prompt**, a short description of the movement, for you and simply makes things move. In manual mode, you write the motion prompt yourself, describing how the subject should move and how the scene should develop.

**Low or high motion.** Low motion suits calm scenes: the camera barely moves, and the subject moves slowly. Its weakness, Holz admits, is that some clips barely move at all. High motion moves both the camera and the subject, which looks lively but can also produce odd, wonky mistakes.

Here is a worked example. Say you made a picture of a lighthouse on a cliff at dusk.

```prompt
Slow push-in toward the lighthouse. The beam sweeps once across the sea.
Waves break against the rocks below. Clouds drift to the left.
```

With **low motion**, expect the camera to creep forward while the waves and clouds stir. With **high motion**, the camera might swing around the tower and the sea churn harder, with a higher chance of a glitch. If one of the four clips works, you can extend it, and each extension continues the action for roughly four more seconds.

> **Try it:** Pick one image you already like and animate it twice, once on low motion and once on high motion, with the same short motion prompt. Compare the eight clips and note which details stayed faithful to your picture and which changed.

## Why it matters this week

**It brings video to a large creative community at a low price.** Midjourney says each job costs about the same as upscaling an image, or roughly one image's worth per second of video, which it describes as more than 25 times cheaper than what has been on the market. TechCrunch notes that video generations use up a monthly allowance much faster than images, so heavy users may need a bigger plan.

**It enters a crowded field.** AI video is already offered by OpenAI (Sora), Runway (Gen 4), Adobe (Firefly) and Google (Veo 3). Many of those tools aim at controllable footage for advertising and film. Early V1 clips look somewhat dreamlike rather than photo-real, in keeping with the artistic style Midjourney's images are known for. How it compares on quality is still an open question.

**It is a step in a bigger plan.** Holz writes that Midjourney's long-term goal is AI that generates interactive worlds in real time, where you can move through a 3D space and everything around you moves too. He describes images, video, 3D and real-time speed as building blocks the company plans to release one by one and then combine.

**It arrives under legal pressure.** The launch comes a week after Disney and Universal sued Midjourney, alleging that its image models produce pictures of their copyrighted characters. Holz's post asks users to use the new tools responsibly. Questions about training data and copyright apply to video just as much as to still images.

## How can you try Midjourney's video model yourself?

You need a paid Midjourney plan; the $10 Basic plan is the cheapest way in. Go to the Midjourney website, make an image as usual, and look for the Animate option under it. Choose automatic to see what the model does on its own, or manual to write your own motion prompt, and pick low or high motion.

To animate your own photo or drawing, drag the image into the prompt bar, mark it as a start frame, and type a sentence describing the movement you want. Only upload pictures you have the right to use, and be careful with photos of real people.

Because each video job costs about eight image jobs, plan before you press the button. Start with images that have a clear subject and a simple background, keep motion prompts to one or two sentences, and extend only the clips that are already working. Write down which settings gave you good results; that small log will save you credits as you learn how the model behaves.

## Sources

- Introducing Our V1 Video Model: https://updates.midjourney.com/introducing-our-v1-video-model/
- Midjourney launches its first AI video generation model, V1: https://techcrunch.com/2025/06/18/midjourney-launches-its-first-ai-video-generation-model-v1/