# Midjourney video: how V1 works, prompt examples, length and cost

> Midjourney V1 video turns an image into a 5 s clip you can extend to 21 s, in 480p or 720p, without sound. Motion prompts that work, camera moves, --loop and --end, GPU cost per plan.

URL: https://superduperai.co/blog/midjourney-v1-video/
Updated: 2026-11-02
Published: 2025-01-15

Midjourney makes video by animating an image: you pick a still, write how it should move (or let Midjourney guess), and get a 5-second clip. Each clip can be extended by 4 seconds up to four times, so **the hard limit is 21 seconds**, at 480p (SD) or 720p (HD), with no sound. The model is still **V1**, launched on 18 June 2025; what changed since is loops, end frames, HD, Discord support and smaller batches.

This page is the short version of the official docs plus the motion prompts we actually reuse. Checked against Midjourney's docs and update posts on 2 November 2026.

## Quick facts

| | Midjourney V1 video (November 2026) |
| --- | --- |
| Input | An image (any Midjourney version, or your own upload). No text-to-video |
| Clip length | 5 s; Extend adds 4 s, up to 4 times = **21 s max** |
| Resolution | SD 480p, HD 720p (for example 832×464 SD and 1280×720 HD from a 16:9 image) |
| Sound | None |
| Videos per job | 4 by default; `--bs 1` or `--bs 2` for fewer |
| Video parameters | `--motion low` / `--motion high`, `--raw`, `--loop`, `--end`, `--bs`; `--video` on Discord |
| Plans | All plans in Fast mode; Relax video on Pro and Mega (SD only); HD from Standard up |
| Cost | SD job of 4 clips = 8 GPU minutes; HD = 26 minutes |

## What's new since the V1 launch

The post you may have read here in 2025 described launch-day V1. Since then:

- **25 Jul 2025:** loops (`--loop`), an end frame (`--end`), and video in the Discord bot.
- **7 Aug 2025:** HD mode, native 720p, about 3.2× the cost of SD. Pro and Mega first.
- **13 Aug 2025:** HD for the Standard plan; batch size 1 or 2 (`--bs`) so a test costs a quarter of a full job.
- **Image side:** V8.2 has been the default image model since 24 July 2026. Images from any version can still be animated.
- **Video model:** no V2. Midjourney's update feed has no new video model through October 2026, only fixes (the Start/End Frame drop zones in August 2026, length-cap messages in October).

So "480p only", "web only", "always four videos" and "no loops" are all out of date.

## How to make a Midjourney video

**On the web (midjourney.com):**

1. Make or upload an image. Its aspect ratio becomes the video's: frame the shot in the image, not in the video prompt.
2. Under the image, choose **Animate**. *Auto* writes a motion prompt for you; *Manual* lets you write it.
3. Pick **Low motion** (still camera, small movement) or **High motion** (camera and subject both move, more glitches).
4. Optional: tick **Loop**, or drop a second image into the **End frame** slot.
5. When a clip works, **Extend Auto** (same prompt) or **Extend Manual** (new prompt for the next 4 seconds).
6. Download for Social (optimised MP4), Raw MP4 or GIF.

**On Discord:** upscale an image and press an Animate button, or post an image URL followed by your motion prompt and `--video`. Turn on remix mode to edit the prompt before animating or extending.

## How to write a motion prompt

Midjourney's docs give the switches but no example prompts. These rules come from the docs and from our own tests:

- **Describe only what changes.** The image already holds the subject, light and style. Spend the words on movement: who moves, which way, how fast.
- **One camera move per clip.** "slow dolly in" or "orbit left", not both. A second move belongs in an extension.
- **Name the speed.** *slowly, gently, sudden, in one quick motion.* Without it, low motion often gives almost no movement at all.
- **Use `--raw`** when the result keeps ignoring your text; it tones down Midjourney's own flair so the prompt carries more weight.
- **Faces and hands:** stay on low motion. High motion is where the glitches live.
- **Static camera:** say "camera static" or "locked-off tripod shot" and let the subject move. Timelapse illusions work best this way: the sky and shadows move, the camera doesn't.
- **Test cheap:** `--bs 1` costs 2 GPU minutes in SD instead of 8. Go back to 4 once the motion is right.

Camera words that read well: *dolly in, dolly out, push in, pull back, orbit, pan left/right, tilt up/down, crane up, drone rise, tracking shot, handheld follow, whip pan, slow zoom*. The full list with what each one looks like is in [camera movement vocabulary](/library/camera-movement-vocabulary/).

## Midjourney video prompt examples

All of these are motion prompts: paste them into *Animate → Manual* on top of a matching image. The parameters are the documented video ones. We rewrote the 2025 list here: the old prompts mixed image parameters like `--ar` and `--v 6` into video jobs, and Midjourney strips those.

### Product and branding

Open: https://superduperai.co/library/product-hero-shot-generation/

```text
camera orbits slowly around the bottle, a soft highlight travels across the glass, faint mist drifts at the base --motion low
```

Open: https://superduperai.co/library/product-hero-shot-generation/

```text
the watch rotates a half turn on the plinth, rim light sweeps across the bezel, camera static --motion low --raw
```

```text
sneakers lift off the floor and hover, slow 360 turntable rotation, dust motes floating in the light --motion high
```

### Portraits and characters

Open: https://superduperai.co/library/camera-movement-vocabulary/

```text
slow dolly in toward her face, hair moving gently in the wind, she blinks once --motion low
```

```text
the detective lights a cigarette, smoke curls up through the lamp light, slight push in --motion low
```

```text
the dancer spins and her dress flares out, handheld camera following her --motion high
```

### Landscapes and timelapse

Open: https://superduperai.co/library/nature-documentary-long-lens-prompt/

```text
clouds race across the sky, shadows crawl over the dunes, heat shimmer on the horizon, camera static --motion low
```

```text
drone rises over the forest and reveals the mountain range at sunrise --motion high
```

```text
waves roll in and break on the rocks, clouds drift left, seamless --loop --motion low
```

### Loops for music and backgrounds

Open: https://superduperai.co/library/music-visualizer-abstract-loop-prompt/

```text
endless dolly forward through the kaleidoscope tunnel, colours pulse slowly --loop --motion low
```

Open: https://superduperai.co/library/music-visualizer-abstract-loop-prompt/

```text
the vinyl record spins, light reflections sweep across the grooves --loop --motion low
```

### Food and drink

Open: https://superduperai.co/library/food-macro-close-up-video-prompt/

```text
chocolate syrup pours slowly over the ice cream, macro push in, glossy drips --motion low
```

Open: https://superduperai.co/library/food-macro-close-up-video-prompt/

```text
steam rises from the coffee cup, morning light shifts across the table, camera static --motion low --raw
```

### Transitions with an end frame

Open: https://superduperai.co/library/image-to-video-first-frame-method/

```text
the city lights switch on one by one as dusk falls --end {URL of the night version of the same image}
```

```text
the empty room fills with furniture piece by piece, camera static --end {URL of the furnished room}
```

### Story in four beats (Extend Manual)

Start with a 5-second clip, then extend three times with a new prompt each time. Each extension starts from the last frame of the previous one.

```text
1. the cat sits on the windowsill, tail flicking, camera static --motion low
2. the cat crouches and wiggles, ready to jump --motion low
3. the cat leaps off the sill toward the table --motion high
4. the cat lands, knocks over a cup, looks at the camera --motion high
```

For the image side, the [first-frame method](/library/image-to-video-first-frame-method/) explains how to build a still that animates well, and [style lock](/library/style-lock-block/) keeps a series of stills consistent.

## Length, resolution and cost

**Length.** 5 seconds per generation; Extend adds 4 seconds, four times at most: 5 + 4 × 4 = 21 seconds. Each extension costs the same GPU time as a new job. If you need longer, cut several clips together.

**Resolution.** The video's shape follows the image:

| Image aspect | SD | HD |
| --- | --- | --- |
| 1:1 | 624×624 | 960×960 |
| 2:3 | 512×768 | 784×1168 |
| 16:9 | 832×464 | 1280×720 |
| 1:2 | 448×880 | 672×1360 |

There is no 1080p. Midjourney does not publish a frame rate; third-party pages say 24 fps, which we have not measured.

**GPU minutes per job** (from the docs):

| | 4 clips | 2 clips | 1 clip |
| --- | --- | --- | --- |
| SD | 8 min | 4 min | 2 min |
| HD | 26 min | 13 min | 7 min |

**Plans** (monthly price; annual billing is 20% off):

| Plan | Price | Fast GPU | Video |
| --- | --- | --- | --- |
| Basic | $10 | 3.3 h | SD, Fast only |
| Standard | $30 | 15 h | SD and HD, Fast only |
| Pro | $60 | 30 h | SD and HD; unlimited SD in Relax |
| Mega | $120 | 60 h | SD and HD; unlimited SD in Relax |

Worked example: Basic's 200 fast minutes are 25 SD jobs, which is 100 five-second clips at the default batch of 4. Extra fast time is $4 per hour on any plan. There is no free tier.

## Midjourney vs other video models

Midjourney's strength is the look of its images and the price if you already pay for a plan. Its weak spots are the ones people search for: no sound, no text-to-video, 720p at most and 21 seconds.

| Model | Sound | Length | Notes |
| --- | --- | --- | --- |
| Midjourney V1 | No | 5 s, extend to 21 s | Image-to-video only |
| Google Veo 3.1 | Yes (dialogue, effects, ambience) | 4, 6 or 8 s; scene extension to a minute or more | See our [Veo 3 prompt guide](/blog/veo3/) |
| Kling 4.0 | Not stated at launch | Up to 30 s in one pass | Launched 15 Oct 2026 |
| MiniMax H3 (Hailuo) | Yes, stereo | 4–15 s | See [Hailuo](/blog/hailuo-ai/) |
| Seedance 2.5 | Yes | Up to 30 s | See [Seedance](/blog/seedance-ai/) |

OpenAI's Sora app closed in April 2026 and its API in September, so the old "Midjourney vs Sora" comparison no longer applies.

A common workflow: make the stills in Midjourney, animate the best ones there for mood shots, and use a model with sound (Veo 3.1) for anything with dialogue. Add music and captions in an editor; our [AI video generation](/products/video/) does this end to end.

## Troubleshooting

- **Nothing moves.** Low motion sometimes gives a near-still clip. Add a speed word and a concrete action, or switch to `--motion high`.
- **Too chaotic, faces melt.** Back to `--motion low`, one action, camera static. The [fix warped hands and faces](/library/fix-warped-hands-and-faces/) recipe has more.
- **The loop jumps.** Use `--loop` instead of asking for "seamless" in words; circular motion (orbit, spin, swirl) hides the seam best.
- **The prompt is ignored.** Add `--raw`, cut adjectives, keep the verb first.
- **Image parameters vanished.** Expected: video jobs accept only the video parameters listed above.

## FAQ


### How long can a Midjourney video be?

5 seconds per generation. You can extend a clip by 4 seconds up to four times, so the maximum is 21 seconds. Longer videos have to be edited together from several clips.

### Is Midjourney video free?

No. Video needs a paid plan; the cheapest is Basic at $10 a month, which can make SD video in Fast mode. A default SD job of four clips uses 8 GPU minutes.

### Does Midjourney video have sound?

No. V1 outputs silent clips. Add music or voice in an editor, or use a model with native audio such as Veo 3.1.

### Can Midjourney make video from text only?

Not directly. V1 animates an image. Generate an image from your text first, then press Animate.

### How do I control the camera in a Midjourney video?

Write one camera move in the motion prompt (dolly in, orbit, pan, tilt, crane up) with a speed word, and choose low motion for a steady camera or high motion for big moves. For a still camera, write "camera static".

### Does --style raw work for video?

Video jobs use their own `--raw` parameter, which reduces Midjourney's added style so your prompt has more effect. Image parameters on the start image, `--style raw` and `--ar` included, are stripped from the video job.

### Is there a negative prompt or an API for video?

There is no `--no` among the video parameters, and Midjourney has no official public API. Services that sell a "Midjourney API" are unofficial wrappers.


## Sources

Checked on 2 November 2026:

1. [Midjourney docs: Video](https://docs.midjourney.com/hc/en-us/articles/37460773864589-Video) (parameters, lengths, resolutions, GPU cost)
2. [Midjourney docs: Comparing plans](https://docs.midjourney.com/hc/en-us/articles/27870484040333-Comparing-Midjourney-Plans)
3. [Introducing our V1 video model](https://updates.midjourney.com/introducing-our-v1-video-model/) (18 Jun 2025)
4. [Looping and end frame, video in the Discord bot](https://updates.midjourney.com/looping-and-end-frame-for-video-video-in-the-discord-bot/) (25 Jul 2025)
5. [HD mode for video](https://updates.midjourney.com/hd-mode-for-video/) (7 Aug 2025)
6. [New video options](https://updates.midjourney.com/new-video-options-and-a-home-for-moodboards/) (13 Aug 2025)
7. [Midjourney docs: Version](https://docs.midjourney.com/hc/en-us/articles/32199405667853-Version) (V8.2 default)
8. [Google: Veo 3.1 in the Gemini API](https://developers.googleblog.com/introducing-veo-3-1-and-new-creative-capabilities-in-the-gemini-api/)
9. [The Decoder: Sora shutdown dates](https://the-decoder.com/openai-sets-two-stage-sora-shutdown-with-app-closing-april-2026-and-api-following-in-september/)
10. [Kling 4.0 press release](https://bernama.com/en/press/news.php?id=2622471)
