Content Creation for Immersive Digital Services

Role: Team member with Dora Yue Du and Lisa Thahn | Duration: 2026 | Focus: Video Narrative Design, AI-Generated Content, Immersive Media Production

This course produced two connected deliverables: a full video narrative treatment for a short film called "Porto Through Borrowed Eyes," and a visual learning diary documenting hands-on experimentation with AI image and video generation tools.

The narrative treatment structured a video around three Erasmus students experiencing Porto for the first time, each carrying a different emotional register through the same day. The film opens with a sequence built entirely on sound before sight: close-ups of hands sorting market fruit, feet on cobblestones, an ear listening to the city, with no faces revealed. From there it moves through three character arcs tied to time of day. Student A carries the morning, defined by expectation and energy, shot with fast cuts and high contrast color through a farmer's market and coffee shop scene. Student B carries noon, defined by discovery and belonging, with warmer tones and slower pacing built around shared meals and river views. Student C carries the sunset, defined by freedom and flow, filmed with long takes and natural golden light on a beach volleyball scene. The film closes on a triptych split screen bringing all three perspectives together, then a single shot of all three walking into the sunset as the sound design strips down to waves alone. Each section paired specific visuals, sound design, mood notes, and voiceover lines, along with a stated narrative function, tracking how Porto shifts from something external and observed to something relational, then emotional, then simply lived in. The treatment also included practical production notes: equipment (gimbal, super zoom lens, external audio recorder), lighting strategy tied to natural light and golden hour, and post production choices like native language narration with yellow subtitles.

The visual learning diary documented a structured experiment across two categories of generative AI tools. For images, Adobe Firefly running the Gemini 3.1 Nano Banana 2 model was used to generate seven Porto themed images from prompts like "cinematic Porto street at golden hour," comparing realistic against stylized outputs and testing how prompt wording shifted the result. For video, Runway ML Gen-4 Turbo was used in an image to video workflow: each AI generated still was fed back in with a motion prompt describing camera movement and mood (a slow push forward through a street, an aerial drift over a bridge at night, rain falling on azulejo tiles), producing three 5 second clips at 16:9.

The reflection section is where the real learning shows. The most difficult part was learning that motion prompt language is a distinct skill from image prompt language, words like "slow push," "aerial drift," and "dolly shot" changed the camera movement outcome significantly, and Runway's interface required finding the correct image to video mode rather than its general tool. What worked well was Firefly's consistency on architectural detail, azulejo tiles, cobblestones, and tram lines all read as convincingly Portuguese, and Runway's ability to generate smooth cinematic motion from a single static image. The ethical reflection was the most substantive part of the diary: these images are photorealistic but entirely synthetic, they depict a Porto that was never photographed, and presenting them as real would be misleading. The diary concluded that AI generated content lowers the barrier to polished visuals dramatically, useful for prototyping and learning, but it cannot yet substitute for footage with genuine location presence, particularly given the occasional visual inconsistencies still present in generated clips (the cloud time lapse being one example).

Skills Demonstrated: narrative structure and treatment writing, shot list and sound design planning, AI image generation (Adobe Firefly, Google Gemini Nano Banana 2), AI video generation and prompt engineering (Runway ML Gen-4 Turbo), image to video workflow design, and critical evaluation of AI generated media ethics and disclosure.

Details of the full narrative treatment and visual learning diary can be found here: Google Drive