
Inspiration
Jude Chiu, The Last Aquarium
What if the animals in the songs are humanized and humans turn out to be the exhibits and consumables? An imagination of the power reversal between humans and marine animals.
Production
ChatGPT
Since ChatGPT is good at forming ideas and stories, I communicated with it to form the storyline below.
The film unfolds amidst the neon-lit, bioluminescent coral reefs of AquaMetropolis, where humans build their undersea cities alongside marine creatures that curiously observe their efforts. Initially, scenes appear tranquil, highlighting the coexistence between humans and marine life. However, as the narrative deepens, viewers witness the fragility of this equilibrium. One of the most poignant moments in "AquaMetropolis" occurs during a powerful underwater event—an awe-inspiring storm or a captivating marine migration. This event serves as a stark reminder of humanity's vulnerability and powerlessness in the face of nature's unyielding forces. The film captures the shock and humility that humans experience when confronted by the overwhelming beauty and strength of the natural world, urging us to reevaluate our place within it. "AquaMetropolis" is a visually stunning and emotionally charged exploration of a future where harmony with nature is a fragile yet deeply meaningful journey, all set against the backdrop of a Cyberpunk-inspired undersea world.
Midjourney
I used Midjourney for text-to-image generation because it is known for its well-trained aesthetics and various art styles.
Pika Labs and Runway ML
I used Pika Labs and Runway ML for image-to-video generation, and eventually chose Pika Labs because it is free and stable.

Photoshop Generative Fill
I used Photoshop together with Midjourney for image generation, fine-tuning, and image expansion.Photoshop Generative Fill function allows me to specify what to fill in a specific range in the image and comes in handy for expanding the original image. While Midjourney's image-expanding function is usually of fixed size, Photoshop Generative Fill allows image expansion to any dimension required.




Audio & Video Editing
To generate narrated audio, I put the transcript into Elevenlabs, a text-to-speech AI tool. ElevenLabs provides many voices and allows voice customizations including Stability, Clarity+Similarity Enhancement, and Style Exaggeration.
I used Stable Audio to generate the BGM and it worked better than my expectation. I used MyEdit for generating sound effects like the explosion sounds.
For video editing, I tried Descript, Premiere Pro, and CapCut. I found CapCut most handy and efficient for my project.