This project tested a fully web-based GenAI workflow using Google-powered tools through Adobe Firefly, specifically Nano Banana and Veo 3.1. The goal was not just to create polished standalone frames, but to see whether an online platform could hold together a true multi-shot sequence with consistent character fidelity, controlled shot progression, and believable human-animal interaction inside one stylized beach narrative. At its core, this was a test of continuity, sequencing, and visual control across an entire piece rather than isolated moments.

The entire background track was generated using Suno, making this ad fully AI-produced end to end.

Nano Banana produced glossy, commercially strong images, especially for beauty-driven shots and controlled editorial compositions, but the workflow was consistently slowed by heavy guardrails. Anything near visible skin, face-adjacent edits, or prompts that resembled identity manipulation introduced friction quickly, leading to repeated prompt rewrites, rejected generations, and corrective passes that interrupted the creative pace. That made the process feel less like directing and more like negotiating with the system. By comparison, offline tools such as LTX and WAN require more hardware, setup, and upfront cost, but they offer something more valuable in experimental storytelling: freedom. Less convenience, more control. And for sequence-building, that tradeoff can matter more than a polished interface.