How to Make a Video From Midjourney Architecture Images: Animate Button or a Six-Image Film

You have a Midjourney image of a building you like — a timber house on a slope, a concrete library, a courtyard with one good tree — and you want it to move. Midjourney's own Animate button is the obvious first step, and for a five-second clip it works. A 30-second vertical film of the same building is a different job: you need six views that agree with each other, a fixed 9:16 frame, sound, and an honest label. This guide covers what Midjourney Video does as documented, how to hold one building across six images, what warps, and when to leave Midjourney for the film.

Key takeaways

• Midjourney's Animate button turns one image into a five-second clip, extendable to a documented maximum of 21 seconds, at 480p or 720p.
• The clip's shape follows the starting image; the documentation lists no 9:16 frame and mentions no audio, so vertical social video needs editing afterward.
• To keep one building across six images, attach the hero as an Edit model reference, reuse a fixed description block, and check bays and roof lines yourself.
• Low Motion is the safer setting for buildings; High Motion is documented as more likely to produce glitchy movement.
• A six-image film from a tool such as Arch2Video gives you a 1080 × 1920, roughly 30-second cut with a soundtrack and no prompting, in exchange for control over each camera move.

The short answer: a clip in Midjourney, a film from six images

To make a video from Midjourney architecture images, open the image on midjourney.com, use the Animate button for a five-second clip, and extend it in four-second steps up to 21 seconds. For a full 30-second vertical film, generate six consistent views of the same building in Midjourney — hero exterior, approach, a second elevation, two interiors, one detail — and either stitch Midjourney clips of each in a video editor or upload the six images to a tool that plans and cuts the film for you. Either way, present the result as concept work, not a built project.

What Midjourney Video gives you, as documented (as of October 2026)

Midjourney's Video documentation describes the feature as turning "a single image into a dynamic 5-second video," with an optional text prompt. What matters for architecture, as of October 2026:

• The buttons. Every gallery image has Animate Image buttons: Animate Manually (edit the prompt first), Auto, and Loop (end frame matches the start). Uploaded images work too, but Style References, Image Prompts and Edit Model references are listed as not compatible with video.
• Length. Five seconds to start; each extension adds four seconds, up to four times, "until it reaches 21 seconds, which is the maximum length available," and costs the same GPU time as the first generation.
• Motion. Low Motion, the default, is "more likely to result in still scenes, low camera motion, slow motion, or subtle character movements"; High Motion "may also create unrealistic or glitchy movements." --raw reduces Midjourney's own additions so your prompt carries more weight.
• Resolution and shape. 480p by default; Standard, Pro and Mega plans can choose 720p. Output size follows the starting image (a 1:2 image gives 672 × 1360 at HD), and Midjourney "may need to adjust the aspect ratio slightly."
• Cost. Each prompt makes four clips by default; the documentation lists four SD clips at 8 GPU minutes (2 for one) and four HD clips at 26 (7 for one). Use --bs 1 while testing.

The plan comparison page lists, as of October 2026, Basic at $10, Standard at $30, Pro at $60 and Mega at $120 per month, with Fast GPU allowances of 3.3, 15, 30 and 60 hours. By those numbers, one 21-second SD clip at batch size one uses about 10 GPU minutes.

Six views of one building in Midjourney: reference tools and a prompt pattern

A film needs the same building six times, not six nice buildings. Midjourney has three reference mechanisms; only one is about geometry.

Style Reference (--sref) holds the look. Midjourney's Style Reference page says it "doesn't copy objects or people, just the overall style — like colors, medium, textures, or lighting": right for one palette across six images, wrong for the same number of bays on the facade.

Edit model references hold the subject. As of October 2026, Midjourney's Version page lists V8.2 as the default, with an Edit model that "replaces Omni Reference, Character Reference, and the Retexture tool." The Edit model page describes generating new images from up to four reference images and has a "Changing Perspective" section with the example prompts "let's see this image from the front / let's see this image from behind" — the documented route to a second elevation of a building you already have.

Image Prompts (--iw) sit in between, and the Image Prompts page says Midjourney uses references "to guide new creations, not to copy them exactly" — a strong suggestion, not a lock.

A workflow that uses them together:

1. Make the hero first and reroll until massing, material and roof form are right, at --ar 9:16 if the film will be vertical (Midjourney's aspect ratio docs list 9:16 for images).
2. Write a fixed description block and reuse it word for word — "two-story house, charred-timber cladding, standing-seam zinc roof, single asymmetric pitch, full-height glazing to the south, late-afternoon sun from the west" — changing only the view phrase.
3. Attach the hero as an Edit model reference for the five remaining views: approach, opposite elevation, main living space, a second interior, a material detail. Add the previous interior as a second reference for the next so the floor and window module carry across.
4. Add the hero as a Style Reference too, so light and palette match even when the Edit model drifts.
5. Check with a pencil, not a feeling. Count bays, compare roof pitch, confirm the detail's materials appear on the exterior, and reroll what fails rather than hoping motion will hide it.

The six-image consistency checklist from the companion article on ChatGPT images applies unchanged, as does the six roles a project set should fill. If Midjourney was only the retexturing step on a Vectorworks, Enscape or Revit render, what to fix in a still before you animate it covers the handoff.

Animating inside Midjourney: what holds up and what warps

Midjourney's video model isn't reading plans; it infers motion from pixels. What survives best is what needs the least invention.

What holds up: a near-static facade on Low Motion, where the camera drifts slightly and clouds move; an interior with one light source and a slow push toward a window; looping hero shots, where the end frame pulls drift back to your image; and atmosphere such as rain on a courtyard or one person crossing a room.

What warps: thin repeated elements — mullions, balustrades, louvers, brick coursing — bend first when the camera slides sideways; a pan or pull-back invents the rest of the street, and that invention won't match your other five images; lettering turns into smears; and each extension starts from the end of the previous clip, so errors compound and a 21-second clip is rarely as clean as its first five seconds.

Write the prompt the way you'd brief a camera operator: one move, a speed, nothing about the building that isn't already in the image. AI video prompts that keep buildings straight ranks the moves by geometric risk, and the ranking holds inside Midjourney: static and slow push-in are safe, pans and tilts invent the edges, orbits invent whole elevations. Keep --raw on so the instruction isn't overridden by a flourish.

Clip or film: choosing by where it will be posted

Arch2Video is the example on the film side: you upload exactly six images from one project (Midjourney renders, drawings or photographs; JPG, PNG or WebP) and receive a 1080 × 1920 MP4 of about 30 seconds. There's no prompt box; an AI director plans the scene order, which may differ from your upload order.

Where it will be posted decides the column:

• A Story, a looping website hero, a one-shot teaser: Midjourney Animate. One image, one clip, Loop on, done in minutes.
• A Reel, Short or TikTok that tells the project: a six-image film. Thirty seconds, vertical, with sound and consistent views is what those feeds reward, and Midjourney's documented limits don't reach it in one generation.
• A client presentation with a specific camera path: stitch Midjourney clips yourself, one per view on Low Motion with matched lighting. Expect light and color to jump between independently generated clips, and budget time to regenerate the two or three that drift.

A six-image film usually takes an hour or more to generate (you're emailed when it's ready), so it isn't for a same-hour post, and it takes your hand off the camera: you review what the director decided.

Vertical output, sound and logos: where Midjourney stops

Vertical frame. Midjourney's video table, as of October 2026, lists output sizes for 1:1, 4:3, 2:3, 16:9 and 1:2 starting images, not 9:16, so plan on cropping or padding to 1080 × 1920 in an editor, and note the source is 720p at best: a 672 × 1360 HD clip has to be scaled up about 1.6× to fill a 1080-wide frame. A six-image film is 1080 × 1920 from the start.

Sound. The Video page describes MP4 and GIF downloads and says nothing about audio, so treat a Midjourney clip as silent and add music in your editor; a studio posting its own work is commercial use, which rules out most trending tracks, and licensed music for architecture videos lists safe sources.

Logo and privacy. Midjourney has no logo overlay; that's an editor step. The plan page also says Stealth Mode, "which keeps your images and videos private," is only on the Pro and Mega plans, so on Basic and Standard a client's unbuilt project sits in the public gallery. An Arch2Video film can carry a studio logo and also comes in a no-logo version.

Disclosure for concept imagery

A Midjourney building is a concept image: nothing was modeled or measured, and the video made from it is an interpretation of an interpretation. Say so where it's posted.

Two rules as of October 2026. Meta's AI labeling policy states: "We'll require people to use this disclosure and label tool when they post organic content with a photorealistic video or realistic-sounding audio that was digitally created or altered, and we may apply penalties if they fail to do so." In the EU, Article 50 of the AI Act requires deployers of systems that generate or manipulate image, audio or video resembling real persons, places or events to "disclose that the content has been artificially generated or manipulated," applying from 2 August 2026; whether an unbuilt house counts is a question for your lawyer, but the label costs nothing.

A caption that does the job: "Concept study generated with Midjourney and animated with AI. Unbuilt; not a construction document." Disclosure wording and platform labels for AI renderings has longer variants and the label steps per platform.

On rights, Midjourney's commercial use page says subscribers own the images and videos they create, even after canceling, with two exceptions: upscaling another user's image doesn't make it yours, and a business grossing more than $1,000,000 a year needs a Pro or Mega plan. Whether an AI-generated image can be copyrighted varies by country, so don't promise a client exclusivity you can't deliver.

Frequently asked questions

Can Midjourney make a 30-second vertical video?

Not in one generation, as documented as of October 2026: Midjourney's Video page caps a clip at 21 seconds, generates at 480p or 720p, and sizes the output from the starting image, with 2:3 and 1:2 listed but no 9:16. Either stitch several clips in an editor and crop to 1080 × 1920, or hand six consistent images to a six-image film tool that outputs 1080 × 1920 at about 30 seconds.

How do I keep the same building across six Midjourney images?

Generate the hero first, then attach it as an Edit model reference (up to four images in V8.2) for each further view, reusing one fixed description block and changing only the view phrase. Add the hero as a Style Reference so light and palette stay matched, then check bays, roof pitch and materials yourself; Midjourney describes references as guidance rather than exact copies, so expect a few rerolls per view.

Can I use Midjourney images commercially in a client video?

Per Midjourney's commercial use page as of October 2026, subscribers own what they create and may use it commercially, except that upscales of other users' images stay theirs and businesses grossing over $1,000,000 a year need a Pro or Mega plan. Your images and videos are public unless your plan has Stealth Mode, and copyright for AI-generated images is unsettled in many countries, so ask a lawyer before promising exclusivity. This isn't legal advice.

Do I need to label a Midjourney-based video as AI?

Yes, in practice. Meta requires its disclosure tool for organic posts with photorealistic AI-generated or altered video and may apply penalties otherwise, and the EU AI Act's Article 50 obligations apply from August 2026 to content that could pass as real. Beyond the rules, a Midjourney building was never designed or modeled; calling it a concept study protects you with clients and planning officers who might read it as built work.