In this project, I wanted to create a short video about a prison guarded by crocodiles, but from the crocodile's perspective.
Images: GPT Image 2.5
Video: Google Omni 1.1, Seedance 2.5
Music: Suno
Compositing (Edit & Color): After Effects, Premiere
The Characters
I designed and characterized each crocodile individually, using GPT image 2.5. Each one has its own distinct physical traits, clothing, accessories, and habitat. Each one even has a dedicated soundtrack that defines him.
After defining the characters, I created a style sheet for each one of them.


Locations
Then I built out each crocodile’s world and location









Scene development
The next step was developing rich scenes for each character, capturing daily routines through close-ups and varied angles.















Reactions
The next step was creating the shots for when things change and the crocodiles receive their summons - the prison envoy delivering the letters, their reactions, etc.












The Prison
I generated the prison from multiple angles, keeping it visually consistent across all shots. It’s located in the middle of the desert, surrounded by a water canal.






Gear Up Montage
For the gear-up montage, I focused on close-ups inspired by 90s action films.






The Storyboard
I created a storyboard in Figma, which allowed me to verify the entire flow and ensure nothing was missing.

Video Generation
For the video generation, I primarily used Google Omni 1.1 (fast and cost-effective), alongside Seedance 2 and 2.5 for specific scenes.
To achieve cohesive multi-shot sequences, I used ChatGPT to structure a JSON prompt for each batch of images, then fed the images as visual references (along with the JSON) directly into Omni 1.1.
Once I had most of the video clips ready, I assembled them in Adobe Premiere.
Music
With a rough cut in place, I wanted unique background music tailored to each crocodile and scene. I used ChatGPT (Astra) to analyze the full video edit and measure the exact duration of each segment. Based on my stylistic direction for each part, ChatGPT generated Suno prompts, which I pasted into Suno manually.
After generating all the audio tracks, I brought them into the project and edited them into the timeline manually.