Project Description (Project Link, Loom)

CUBE is a short cinematic sci-fi teaser set in a future where humanity survives inside massive cube megastructures after Earth becomes environmentally uninhabitable. The idea came from a concept I had imagined for years and always wanted to visualize in a cinematic way. I wanted the world to feel atmospheric, grounded, quiet, and emotionally restrained rather than like a traditional sci-fi action film. The teaser starts with a peaceful vision of Earth and slowly reveals that nature only exists as a projection inside a brutalist megastructure. Everything was built entirely inside Melius using image generation, image-to-video workflows, stitching, refinement, and cinematic worldbuilding.

Process Steps

-Started with a sci-fi concept and visual world I had imagined for years

-Generated multiple images to establish the atmosphere, architecture, lighting, and cinematic tone

-Iterated prompts repeatedly to create visual consistency across the world

-Built moodboards and references before developing the actual teaser shots

-Created the teaser shot-by-shot using connected image generations and image-to-video workflows

-Used generated images as references for later generations to preserve continuity

-Built the opening sequence by stitching together multiple generated clips into one cinematic pullback shot

-Used video-to-video workflows and refinements to smooth transitions and motion

-Stitched sequences together directly inside Melius to make the final teaser feel cohesive

Feedback on Melius

This was my first time working with a node-based workflow, so understanding how different nodes connected together was definitely the biggest learning curve for me. I’m still figuring a lot of it out while experimenting on the fly, especially understanding how different workflows interact with one another inside the canvas.

I think having an in-canvas support or tips system would help a lot, especially for first-time users trying to understand nodes, references, and workflow structures without constantly leaving the canvas. Small contextual guidance while building would make the learning process much smoother.

One workflow limitation I personally struggled with was not being able to attach a reference image while using audio from another clip during refinement or stitching workflows. I’m not sure about the technical limitations behind it, but having that flexibility would have helped a lot while maintaining cinematic continuity across shots.

I also think having a better way to refine and merge audio seamlessly while stitching clips together would make the workflow much smoother for cinematic editing and transitions.

The most rewarding part of the entire process was finally seeing an idea I had imagined for years come to life visually for the first time.