Data Requiem : Self-Directed

ABOUT

Data Requiem is a self-directed short film set in a cyberpunk future where memory, identity and data collapse into one another. It began as a question: can generative AI tools sustain a consistent visual narrative across a sequence of scenes, or does coherence break down?

Initially, this started as me teaching someone how to use Runway then I ended up generating some really cool video pieces in an animated style, then I realised the potential in this, could I make an anime with this?

PROCESS

The film was built using Runway Gen-4 and Midjourney. The central creative challenge was character continuity, maintaining a consistent visual identity across generated sequences without the tools natively supporting it.

It was quite tricky to solve the continuity problem because generations are limited to 5 or less seconds, so you have to be able to get each scene a good enough quality then move to the next. This was a problem because I spent a lot of credits on generations that were really poor in quality so it was a lot of RnD to try and get the right scene. On the bright side, I learned a lot about the capabilities of AI video generation on a budget, I loved the whole process of storyboarding then taking the characters I created and put them into the scenes was really cool, despite the consistency of their appearance not being 100% and you can see this in how the main cyborg character looks slightly different in every scene

OUTCOME

Overall the project was great fun for me experimenting, I was able to teach a lot of people about how to use Midjourney and Runway, and also got featured in the motion for our BCU Graphics Gradshow.