How We Created Parcfilm’s Recruitment Video for a Junior AI Creative
The video was designed as a blend of live-action footage, AI generation, and traditional post-production compositing. The idea was to turn a recruitment announcement into a video that immediately conveys the way we work.
The final result started with a controlled live-action shoot and then went through a workflow in which each fictional element was created separately. This made a significant difference because when you try to generate too many elements at once, quality drops quickly and unwanted artifacts become difficult to fix.
We figured that out pretty early—after burning through our first AI credits on generations where we asked for every effect to happen all at once, within the same prompt.
Wrong.
How the Workflow Was Designed
After cleaning up the audio and color grading the footage in DaVinci Resolve, we captured a screenshot from the widest shot—the frame showing the largest view of the scene. This frame became the foundation for generating our AI references.
Using that image in Magnific, we generated the fictional characters as visual references: a wolf created entirely from scratch and replacing our editor colleague with a pink unicorn. The purpose of these images was not just to look good, but to establish consistent visual references for the next stages of the video generation workflow.
Once we had this foundation, we exported the three video clips separately, corresponding to the three different camera angles. We then uploaded the wide-angle clip along with the previously generated image of Tudor surrounded by the fictional characters into Magnific Video Generator to create the first AI-generated video version.
The Correct Generation Order
The first step was replacing the woman with the pink unicorn through a dedicated generation. Once that element was completed, we ran a second generation using the same video as the base, this time introducing the wolf.
The order mattered a lot because each new character increased the risk of the model losing consistency and distorting the rest of the frame. Based on what we learned during this project, the best approach was to build the scene layer by layer rather than trying to generate everything with a single prompt.
After generating the widest shot, we captured a screenshot and used it as the reference for the remaining camera angles. This kept all perspectives visually consistent, ensuring that the fictional characters maintained the same appearance and visual logic across every shot.
How the Birds and Lighting Effects Were Added
The birds were generated separately using the original footage, then composited onto the AI-generated versions afterward. In this case, the best solution was to add them in After Effects using Rotobrush, as generating them directly through the prompt produced noticeable artifacts.
This is where one of the most useful workflow principles comes into play: if a secondary element compromises the main generation, don’t force it into the prompt. It’s better to add it later using traditional compositing than to compromise the entire sequence.
The exact same approach was used for the lights that turn on, switch off, or change color. These effects were handled separately in post-production because integrating them directly into the AI generation would have made the scene more unstable and much harder to control.
What Worked Best
The most effective workflow was this: first, create visual references for the characters; then generate each one individually; and finally composite the remaining elements in After Effects. This approach provided much greater control over the final result while significantly reducing the risk of artifacts.
It also helped preserve the artistic consistency of the video. When you’re working with multiple camera angles, several fictional characters, and lighting effects, the production order becomes almost as important as the creative idea itself.
About the Sound
The sound effects were added and synchronized manually by the editor rather than generated with AI. While AI sound generation can produce excellent results in many situations, for this project it was more reliable to maintain full manual control over the audio.
This also helped refine the video’s final timing, because the sound should support the visuals, not compete with them. In a recruitment video, clarity and pacing are more important than showing off technical capabilities.
Conclusion
This video is a good example of how AI can be used in video production—not as a complete replacement for post-production, but as a tool for accelerating workflows and enabling creative experimentation. When the references are built properly and the generation process is approached step by step, the final result offers greater control with fewer compromises.
For Parcfilm, the goal was twofold: to create a recruitment video that stands out from the crowd while also demonstrating, in a practical way, how we approach hybrid production—combining live-action filming, AI, and traditional compositing.