Tim Simmons revisits an earlier AI short film to compare a manual production workflow with an agent-assisted remake. Tim Simmons uses structured folders for story, characters, locations, references and shot tracking, then asks GPT-6 Astra to derive a script and shot list from the previous film. Character turnarounds supply additional visual references, although Tim Simmons labels their consistency benefit as his own hypothesis.
Tim Simmons builds a local production dashboard to make a dense project folder understandable. The dashboard connects shots, prompts, storyboards and session notes so both the filmmaker and the agent can recover what happened between sessions. Voice instructions guide generation through a tool connection, but Tim Simmons still reviews the returned clips and approves or rejects them.
Tim Simmons describes waiting between prompting, generation and file organization, and finds that manual editing remains faster in his experiment. Agent assistance becomes more useful for a color pass, small timeline adjustments, titles and audio leveling in DaVinci Resolve. Tim Simmons distinguishes grading existing rendered footage from recovering raw image information and emphasizes human oversight, tool limitations and the cost of unnecessarily expensive reasoning.
Watch on YouTube
