Project Link: https://app.melius.com/projects/3db4486f-0982-454d-bab7-c2459bf781e3/canvas/7a183053-2837-463e-a0dd-0d1056e9f57e

I built this crazy hyper-kinetic ad campaign for the "AURA-X" techwear sneaker. I went all in on the dark liquid chrome, glowing cyan energy, and insane fluid dynamics. I mapped the whole pipeline out in Melius Canvas—starting with Gemini Banana 2 to lock the exact base assets, then pushing them directly into Veo 3.1 and Kling 3 Pro to get that relentless, aggressive video momentum. We created 7 completely different scenes, from deep-sea drops and melting tube waves to a mid-air skydiving chase and a sick liquid metal logo reveal. It's totally cinematic and high-end.

* Process Steps :

Step 1: The Base Assets: Started with Gemini Banana 2. I used it to lock the sneaker's design (matte black, liquid chrome cage) and generate the exact Start and End frames. Locked the 30-degree angles and lighting perfectly.

Step 2: Hyper Motion Setup: Connected those image nodes straight into Veo 3.1 and Kling 3 Pro. I used these models because I wanted non-stop camera movement—whip pans, orbits, and speed ramps with the shoe dead center, and they handle that kinetic energy perfectly.

Step 3: The Separate Clip Fix: When I tried doing the big elemental transition (ice to fire to wind), the video engines were losing coherence across the full 8 seconds. So I divided it into three separate 3-second match-cut nodes. I kept the camera angle exactly the same in the Gemini Banana 2 image prompts, which made it seamless to stitch together in the edit.

Step 4: Bypassing Safety Filters: On the airplane scene, the "shattered glass" prompt hit an AI safety block. I just tweaked the prompt to an "open side hatch" with atmospheric vapor. Kept the high-fashion energy without triggering the filters.

Step 5: The Outro: Ended it with a liquid metal vine grabbing the shoe, then ran a text-to-video node using Kling 3 Pro / Veo 3.1 to perfectly trace the AURA-X logo in glowing chrome.

*.My Feedback for Melius: Honestly, the node-based architecture is a massive upgrade. It totally fits my vibe coder workflow. Being able to visually pipe the start/end frames directly into the video nodes gives me insane control over the art direction. When a longer video generation hallucinates or breaks down, it's so easy to just split it into smaller micro-batches on the canvas and fix the issue.

The only real friction was hitting those opaque safety filters on some of the more creative action shots. But because of how fast Melius is, duplicating a node, fixing the text, and reconnecting the edge takes almost no time. It totally bridges the gap between just writing prompts and doing real, high-level technical direction.