Midjourney's first video model: what image-to-video delivered at launch, its honest limits, and where it lands in a 2026 market that Sora exited entirely.
π 2026 update: This is an archived review of Midjourney's Video Model V1, launched in the V7 era (2025). Video is no longer a separate experiment β it's a core capability: clips now run 5 to 21 seconds with materially better motion coherence under V8.1+. For today's cross-vendor picture, see the AI video comparison. We keep this review live as the record of where Midjourney's video line started β useful context for judging how fast it's moving.
Video Model V1 did one thing β animate Midjourney stills β and did it with the platform's signature aesthetic intact. Short clips, simple motion, frequent artifacts on complex movement; a promising toy with a real superpower: your Midjourney look, moving.
The bet paid off in its own lane. Midjourney never chased Veo and Runway into cinematic text-to-video; it built the fastest path from a still you love to a living visual β and for social, concept and mood work, that's frequently the whole job.
The model that chased generality is gone. The one that stayed in its lane is still shipping. That isn't a coincidence, and the economics below explain why.
Image-to-video animation of generated or uploaded stills β the aesthetic carried over, which was the entire point. You weren't describing a scene from scratch; you were putting an image you had already approved into motion.
Access sat inside the existing workflow. No new tool, no new prompt language, no separate subscription. For an existing Midjourney user, video was a button rather than a migration.
| Then (V1, V7 era) | Now (V8.1+, July 2026) | |
|---|---|---|
| Clip length | Short experimental | 5β21 seconds |
| Motion artifacts | Frequent | Materially improved |
| Product position | Separate-feeling experiment | Integrated core capability |
| Promise | Stills aesthetic, moving | Same promise, more deliverable |
What didn't change: the positioning. Midjourney video remains stills-first animation, not a cinematic generator.
Every constraint downstream traces to this. A composition with a clear subject, uncluttered background and unambiguous depth animates cleanly. A busy frame with overlapping subjects produces the warping people blame on the model. The image is 80% of the outcome.
The instinct carried over from image prompting β describe everything β is wrong here. The still already carries the scene. Describe only what moves, and describe it conservatively. "Slow push in, drifting mist" outperforms a paragraph of cinematic language almost every time.
OpenAI deprecated Sora 2 in spring 2026, with the API shutting down September 24, 2026. The reported economics are brutal and worth stating plainly:
That is not a product that lost a feature race. That is a product whose unit economics never closed. Every vendor still standing is standing because their cost per second works.
The ranking shuffled harder than most people expected:
| Model | Approximate cost |
|---|---|
| Veo 3.1 Lite | $0.05/second |
| Kling 3.0 | $0.07/second |
| Kling 3.0 vs Sora | 65% cheaper |
| Kling 3.0 vs Runway | 44% cheaper |
When generation costs fall this fast, "best model" stops being the deciding question for high-volume work. Cost per usable second takes over.
Midjourney video doesn't appear in the top-10 rankings because it isn't playing that game. Veo 3.1 owns native 4K with audio; Runway owns the production workflow; Seedance and Kling own price-performance at volume.
Midjourney owns something narrower and more defensible: "this exact image, alive, in 90 seconds."
Every other vendor competes on fidelity, length, resolution and price. Midjourney competes on a look people already chose. If your visual identity is Midjourney-built, no leaderboard position changes the fact that only Midjourney animates your images in your style.
The constraint that defined V1 still defines the category: slow beats fast. Push-ins, gentle parallax, drifting particles, cloth and hair movement, atmospheric drift β these hold structural coherence because the model only has to interpolate small displacements per frame.
The mental model transfers intact: still first, motion second, restraint over ambition. Prompts written for V1 motion still read correctly. Nothing about the workflow was invalidated.
Length is the headline β 21 seconds is roughly 4Γ the practical V1 ceiling, which moves it from "loop" to "shot." Coherence improved enough that moderate motion is now viable where only minimal motion worked before. The absence of audio did not change, and remains the defining gap.
Veo wins on native resolution, clip length and β decisively β synchronized audio. Midjourney has no answer to audio and hasn't signalled one. If your output needs sound, this isn't a comparison; it's a different product category.
Those models win at volume: price per second, throughput, and text-to-video from scratch. Midjourney doesn't compete on any of those axes. Choose them when you're generating many clips; choose Midjourney when you're animating a specific image you already approved.
Runway owns the production workflow β timeline, editing, team collaboration. Midjourney is upstream of all that: it produces the asset, it does not produce the edit. The two coexist in a real pipeline more often than they compete.
| Need | Reach for |
|---|---|
| Sound | Veo 3.1 |
| Volume at low cost | Kling 3.0 / Seedance 2.0 |
| Production workflow | Runway |
| Your Midjourney image, moving | Midjourney |
Video generation draws on your existing Midjourney allowance rather than a separate per-second bill. For creators already paying for stills, marginal cost per clip is effectively zero within the plan β a materially different economic position from the per-second vendors.
Clip-level variance is much higher than still-level variance. Budget for generating several and keeping one. That ratio, not the generation time, is the real cost of the workflow.
Treat Midjourney video as b-roll and texture, not as shots. It fills atmospheric gaps, produces motion backgrounds and animates key art. Cutting a narrative sequence from it is possible and consistently more work than it looks.
V1 was the case study in scope discipline: while Sora chased cinematic generality and ultimately exited the market, Midjourney shipped the smallest useful version of video and iterated inside its aesthetic moat.
Eighteen months later, one of those strategies still exists.
The lesson isn't "don't be ambitious." It's that a narrow capability with a defensible reason to exist outlives a broad one with unclosed unit economics. Sora was better at more things. It is also gone.
Sources: Midjourney version docs Β· Digital Applied β AI video market after Sora Β· Teamday β Best AI video models 2026 Β· Pinggy β Best video generation AI models Β· PixVerse V8-era review Β· Spectrum AI Lab market context
Originally published July 2025 Β· Substantially revised July 30, 2026
Seedance 2.5 brings 30-second audio-video generation, 50 mixed references and timed editing. Here is what is official, priced and still worth testing.
MiniMax H3 generates 2K video with native stereo audio, publishes per-second pricing, and offers H3-Base weights under a Community License.
Hermes Agent blends persistent memory with self-improving skills, but its security and outcome depend on the backend, approvals, and review you choose.
OpenClaw is a high-capability agent runtime whose value depends on deliberate gateway security, skill review, and operating discipline.
AI music in 2026: Suno v5.5's capabilities and legal defiance, Udio's settlement path, the Warner deal retiring unlicensed models, and creator rules.
The AI video market after Sora's exit β with real pricing: $0.50 to $2.50 per 10-second clip, who leads the rankings now, and how to choose per shot.
The honest 2026 free-tier comparison: Gemini's generous bundle, Claude's quality-first plan, ChatGPT's capped breadth, and when free stops being enough.
Archived review of GPT-5's August 2025 launch: the unified-router design, the 4o backlash, what improved β and how it reads from mid-2026.
