How to Make an AI Film With the Same Character in Every Scene
A creator demonstrates how to build a consistent, 60-second AI short film using Imagine Art's Film Studio, focusing on the importance of pre-production planning, shot lists, and storyboard mode for character consistency. The walkthrough covers the full pipeline from story concept through generation, editing, and export within a single platform. The video is sponsored by Imagine Art.
Summary
The video opens with a short AI film called 'Glitch' and immediately addresses a core criticism of AI filmmaking: most AI films are disconnected clips stitched together rather than cohesive narratives. The creator argues the problem isn't the AI model itself but rather that creators skip the foundational work real directors do — story development and shot planning.
The creator outlines the film 'Glitch,' a 60-second story following a chrome thief robot in cyberpunk Tokyo who steals a glowing briefcase, gets chased across rooftops, jumps through a portal, and lands in medieval countryside. This narrative is broken into four scenes and nine individual shots, each assigned specific camera language: focal lengths (18mm to 50mm), apertures (f1.4 to f5.6), and camera movements (push-ins, tracking shots, rack focus). The creator emphasizes this planning takes only about 10 minutes but is the difference between a film and a screensaver.
The walkthrough then moves into Imagine Art's Film Studio, a sponsored tool the creator tested over several days before agreeing to cover. The platform has three modes: single shot, multi-shot (up to five shots in 15 seconds), and storyboard mode. Storyboard mode is highlighted as the key differentiator — the creator inputs a logline, character description, and scene breakdown, and the tool returns nine consistent keyframes with the robot's appearance locked across all of them.
For generation, the creator uses multi-shot mode scene by scene, selecting presets for lens, aperture, camera movement, and focus rather than writing lengthy prompts. Scene one (Tokyo rooftop) generates successfully on the first pass. Scene two (the chase) required a second pass after the first came back static. Scene three (the portal) hit the 15-second ceiling and was extended using the platform's 'Extend' feature, which continues a clip forward in the same visual style. Scene four (the medieval landing) took three passes to get the castle scale right.
All generated clips are then assembled in the platform's built-in timeline editor, where the creator adds a cinematic music track, trims one clip, and inserts a dissolve between the portal and landing shots. The full film exports as a 1080p MP4 in under two minutes, all within a single browser tab.
The creator concludes with an honest assessment: storyboard mode's character consistency, the presets panel, and the in-platform editor are genuine wins, while the five-shot/15-second ceiling on multi-shot is a real limitation. A V2 of Film Studio is noted as shipping June 15th with unspecified upgrades.
Key Insights
- The creator argues that the primary reason most AI films fall apart is not the AI model itself, but that creators skip story development and shot planning — the work real directors spend the most time on.
- The creator claims that storyboard mode in Film Studio solves character drift — the single biggest problem in AI filmmaking — by locking the character's appearance, plating, chest glow, and proportions consistently across all nine keyframes regardless of environment.
- The creator explains that lens and aperture choices are doing emotional work, not just visual work — for example, a shallow f2 aperture isolates the robot, and a rack focus from briefcase to face tells the story beat without any dialogue.
- The creator demonstrates that the 'Extend' feature allows pushing past the 15-second multi-shot ceiling on a single clip while maintaining the same visual style, light, lens, and motion physics — making the result read as one continuous shot rather than two spliced clips.
- The creator notes that scene four's medieval castle landing required three generation passes because the first two placed the castle too close, making it look like a setpiece; the fix was rewriting the environment prompt with 'distant silhouette' and 'purple dusk sky' to achieve correct scale.
Topics
Full transcript available for MurmurCast members
Sign Up to Access