Post-mortems, production decisions, and things learned the expensive way - from three finished AI films.
All posts draw from real production on Machine Dreams projects.
The most common failure in AI character generation isn't a technical problem. It's a description problem. The model gives you exactly what you asked for - and what you asked for was a type, not a person.
Two bad results in a row don't mean the third will be better. They mean the prompt is broken. The distinction matters more than it sounds - because the reflex is always to try again.
The place had to feel real before anyone appeared in it. A deliberate choice that shaped every subsequent decision in the production.
Before there was a script, I designed the Summit Rest System in full. Reference sheets, material close-ups, technical drawings. Here's why that mattered.
In Between is a film about Tuesday. No journey, no transformation. The challenge was resisting the pressure to make it mean something it didn't.
Every production failure I've experienced came from the same place: starting too early. Not too early in the day - too early in the process.
Could Be Better. Two words that kept the Arctic thriller moving when the alternative was holding production hostage for a 5% improvement that wasn't coming.
Every project I've made started the same way - not with a tool, not with a prompt, not with a shot list. They started with a question. Two words. No technical requirements, no budget implications, no deadline.
All posts on this page draw from a complete production workflow guide - 126 pages, eight chapters, built from three finished films.
GET THE GUIDE - €29 Download free sample - 2 chapters, no cost