How to Turn Midjourney Images Into Videos

Midjourney video works differently than most AI video tools. You do not type a prompt and get a video from nothing. You start with a still image that already exists in Midjourney, then convert it into a short animated clip. This image-to-video approach gives you control over composition and subject before any motion is applied.

This guide covers the full workflow: generating the source image, applying the --video parameter, understanding plan and speed requirements, and getting your first usable video output. Every step uses real prompt examples you can paste into Discord or the Midjourney web interface.

How Midjourney video generation works

Video generation in Midjourney is exclusively image-to-video. You cannot pass a text prompt directly into a video generation. Instead, the process is always two stages. First, you generate or upscale a still image. Second, you apply the --video parameter to that image to create a short animated clip. This is fundamentally different from tools that generate video from text alone.

The output is a short clip, typically a few seconds, where the camera or subject moves based on Midjourney's interpretation of the scene. You can influence the amount of motion, whether it loops, and (in some cases) where the motion ends, but the generation is still guided by what the original image contains.

Plan and speed requirements

Video generation is available across all Midjourney plans when using Fast mode. If you want to generate standard-definition video in Relax mode, you need a Pro or Mega plan. HD video is more resource-intensive and its availability depends on your plan tier and whether you are using Fast hours. Budget your Fast time accordingly if you plan to create many clips.

• Fast mode video: available on all plans (Basic, Standard, Pro, Mega)
• Relax mode SD video: Pro and Mega plans only
• HD video: plan and Fast dependent, uses more GPU time per generation
• Each video generation consumes Fast hours (or Relax queue time on eligible plans)

Step 1: Generate a strong source image

The quality of your video depends almost entirely on the quality and composition of the source image. Choose subjects with clear implied motion or environmental depth. A figure mid-stride, waves hitting a cliff, or clouds over a landscape will animate far better than a flat product shot or a logo.

a lone astronaut walking across a red desert, wide shot, cinematic, golden hour lighting --ar 16:9

ocean waves crashing against volcanic rocks, misty spray, dramatic sky, slow shutter feel --ar 16:9

Use --ar 16:9 for widescreen clips. The aspect ratio of the source image carries into the video output, so set this at the image generation stage.

Step 2: Apply --video to the image

Once you have upscaled your image (or selected a variation you like), you apply the --video parameter. In Discord, this is done via a reaction or by referencing the image in a new prompt. In the web interface, the video option appears on upscaled images.

/imagine prompt: [image_url] --video

This tells Midjourney to animate the referenced image. You do not need to add descriptive text when converting an existing image, but you can include motion parameters alongside --video to control behavior.

Step 3: Control motion intensity

By default, Midjourney decides how much motion to apply. You can override this with the --motion parameter followed by low or high. Low motion creates subtle movement like gentle camera drift or slight environmental changes. High motion creates more dramatic movement but can introduce artifacts if the scene is complex.

/imagine prompt: [image_url] --video --motion low

/imagine prompt: [image_url] --video --motion high

Start with --motion low for your first experiments. Low motion produces more reliable results and avoids distortion on faces or fine details. Move to high only when you want dramatic camera sweeps or large-scale environmental changes.

Step 4: Add looping and end frames

If you want the video to loop seamlessly, add --loop. This is useful for backgrounds, social media headers, or ambient content where you want the clip to play continuously without a visible cut.

/imagine prompt: [image_url] --video --motion low --loop

The --end parameter lets you provide a second image as the target end frame. Midjourney will animate between the start image and the end image, creating a transition. The --bs parameter controls blend strength between frames.

/imagine prompt: [start_image_url] --video --end [end_image_url] --bs 50

What is not compatible with video

Several Midjourney features that work with image generation do not work with video. Image Prompts (passing a reference image to guide style), Style References (--sref), and Omni References are not compatible with video generations. If you include these parameters alongside --video, they will be ignored or the generation will fail.

• Image Prompts are not compatible with --video
• Style Reference (--sref) is not compatible with --video
• Omni Reference is not compatible with --video
• Personalization (--p) is not compatible with --video
• All style and reference must be baked into the source image before video conversion

Practical examples: full prompts from image to video

Cinematic landscape loop

Step 1: aurora borealis over a frozen lake, wide angle, photorealistic, deep blues and greens --ar 16:9
Step 2: [upscaled_image_url] --video --motion low --loop

Character motion with high intensity

Step 1: a samurai drawing a katana in a bamboo forest, dynamic pose, motion blur on blade --ar 16:9
Step 2: [upscaled_image_url] --video --motion high

Transition between two states

Step 1 (start): a flower bud in morning light, macro photography, dew drops
Step 1 (end): the same flower fully bloomed, warm afternoon light, macro
Step 2: [start_url] --video --end [end_url] --bs 60

Tips for reliable video output

• Compose for motion at the image stage. Scenes with depth, weather, or implied movement animate best
• Lock your aspect ratio early. Use --ar 16:9 for horizontal video or --ar 9:16 for vertical
• Start with --motion low and increase only if the result is too static
• Use --loop for anything meant to play on repeat (backgrounds, ambient content, social posts)
• Keep subjects simple when using --motion high. Complex scenes with many small details can distort
• Budget your Fast hours. Video consumes more GPU time than image generation

Choosing source images that animate well

Not every Midjourney image makes a good video source. The best candidates have implied motion already embedded in the composition. Think about what could physically move in the scene: water, fabric, hair, clouds, particles, vehicles, or the camera itself. If the scene is entirely static (a symmetrical logo, a flat icon, a tightly cropped texture), the video output will feel forced because there is nothing for the system to animate naturally.

• Strong candidates: landscapes with weather, action poses, scenes with depth layers, environments with particles or fluid
• Weak candidates: flat graphic designs, symmetrical patterns, tightly cropped textures, logos, UI mockups
• Test signal: if you can imagine the scene as a 3-second movie clip in your head, it will probably animate well

Common mistakes

• Trying to pass a text-only prompt to --video without a source image. Video requires a starting frame
• Adding --sref or image prompts alongside --video. These are not compatible with video generation
• Using --motion high on detailed portraits. High motion distorts faces and fine textures
• Expecting long-form video. Midjourney video produces short clips, not minutes-long content
• Generating video in Relax mode on a Basic or Standard plan. SD Relax video requires Pro or Mega

Next steps

Once you can reliably turn images into videos, the next level is controlling exactly how that motion behaves. Related reading: /blog/how-to-control-midjourney-video-motion-loops-and-end-frames covers motion intensity, loop tuning, and end frame transitions in detail. For improving the source images that feed into video, see /blog/writing-better-midjourney-image-prompts.