LIVING GAME BOX COVERS

Project: Living Box Art. 80’s Nintendo box art coming to life through Image-to-Video animation in a perfect loop 

Experiment: Stress test the capabilities of Midjourney’s new Image-to-Video on a small scale. 

Tools: Midjourney (image to video), DaVinci (editing). 5 Clips total.

Vision: 5 NES Game Box Covers with a 5 second looping animation, starting with a 2 second pause to absorb the box art “as is”. The animation prompts I used favored whatever natural motion or concept was already present in the box art (i.e. if somebody is climbing, climbing is the focus of the prompt). 

Expectations: I’ve experimented with Midjouney extensively for 4 years, but Image-to-Video is very new for the platform. As with Kling, Nano Banana and homebrew Stable Diffusion; I expected a wide variation in Prompt accuracy and clip viability – and a sizable amount of unusable footage. 

What I Got: And there was a F***-TON of unusable footage! I’d argue that what you see represents barely 5% of the total video clips generated; with the unusable footage being visually flawed, prompt-inaccurate, or just flatly unusable. In addition, it takes time to generate those clips, so 95% of your time is a waiting game for things you ultimately can’t use. 

What I didn’t expect: When it DOES work, it’s a bag of wonders. The Super Pitfall clip in particular genuinely shocked me – though there are fine details in all the clips I find captivating. The perfect looping is a very nice feature, but the Ai slavishly loops at the expense of everything else, even your Prompting! Open-ended clips and Start-to-End Image clips would’ve likely turned out very different.

Results: Midjourney Image-to-Video is, sadly, like many I2V platforms right now: a frustrating roulette wheel with big prizes. Expect 1 viable clip per every 12(!) as a conservative guide, though that ratio can be mitigated with a smaller scope, better prompting, and an open mind. If you’re using Ai for video: don’t ever plan around a vision, plan around surprises. 

🔴If You’re Anti Ai: Position yourself as a Creative that can make surgical changes to abstract creative work. The greatest weakness of Ai right now is that it is a frustrating “I’ll know it when I see it” endeavor. You should promote yourself as a vision-holder that is frustration-free and much quicker than Ai in the long-run. If the Studio has a very clear creative idea of what they want with very little wiggle-room, you can easily argue a Human is the best pick for that job.

🔵If You’re Pro Ai: Ai Video is a bucking bronco without a rider – aimless and wild. You need to position yourself as the ultimate Ai wrangler who can mitigate tons of wasted footage and time with clear, decisive generations. Your “craft” is in making all those precious Fast Hours stretch out as long as possible by spending as little as needed. Pick your poison: free but unstable homebrew solutions or expensive-yet-powerful platforms like Midjourney, Kling, Pixverse, Nano Banana, etc.

All the source files can be found here so you can experiment yourself (right click save as).