Forest Intercept is a 9:16 AI-generated animated short created as part of the Crazy Maple Studio creative test. Unlike my previous self-directed AI-assisted projects, this piece was built around a provided script, with specific constraints on format, runtime, and visual direction. The assessment required a western animation style and a final duration under 30 seconds, which made the project feel more like a focused production challenge than an open-ended concept piece. Because the core narrative was already defined, my creative contribution centered on style development, character design, action choreography, camera direction, music selection, sound design, and editing.

​

For the visual direction, I chose Arcane as a key inspiration. I wanted the film to feel cinematic, stylized, and emotionally intense while still remaining readable within a short vertical format. In the character design process, I focused on creating two visually distinct figures ——Hunter and Stranger, with painterly brush-stroke textures, strong silhouettes, and costume details that could support a Western animated look. For the environment, I leaned into a more 2D illustration-inspired background style, using a dark forest setting with layered fog, cold moonlight, wet ground, tree roots, and high-contrast lighting to create atmosphere and depth.

​

To develop the story further without changing the original script, I used ChatGPT to expand the scene structure and refine the action beats. This helped me define the pacing, character movement, camera language, and emotional tone of the sequence. I then used Nano Banana Pro to generate three-view character references, detail sheets, and style frames for both characters. These images became the visual foundation for maintaining consistency across shots. I also generated environment style frames based on the same visual language so that the characters and background would feel like they belonged in the same world.

​

After establishing the visual direction, I created a complete shot list for the full sequence. This included the number of shots, the duration of each shot, the camera angle, camera movement, which characters appeared in each moment, what actions they performed, how the scene should be staged, and where music or sound effects would be needed. From there, I wrote individual prompts for each shot and generated the video footage using Seedance 2.0. I also tested Kling 3.0 Omni during the process, but the results were not as effective for this stylized animated direction. Based on my previous experience, Kling 3.0 Omni appears to perform better with realistic or photorealistic video styles, while Seedance 2.0 gave me stronger results for this painterly western animation approach.

​

Once I had generated enough footage, I edited the sequence in Premiere Pro, focusing on action clarity, rhythm, and continuity between shots. I also sourced sound effects from Epidemic Sound to support the forest atmosphere, character movement, and fight beats. After the edit was finalized, I used Topaz Video to upscale the final output.

​

This project was both a creative test and a practical workflow exercise. It challenged me to work within a fixed script and production constraint while still making strong creative decisions through visual design, shot planning, and editorial structure. Forest Intercept also reinforced the importance of pre-production in AI-assisted filmmaking. By carefully building character references, environment style frames, shot lists, and shot-specific prompts before generating footage, I was able to create a more controlled and coherent final piece. For me, this project represents a focused exploration of how AI tools can be used not just for image generation, but for structured animated storytelling under real production limitations.