Most AI video content online stops at the final cut. Polished output, mysterious origin. We wanted to do the opposite. “How We Made This” is a series where our creative team walks you through real client deliverables, beginning at the brief and ending at the export, with every meaningful decision in between.
For the first installment, our Creative Director Kean Bartelman broke down a project we delivered for Japan Tobacco International — a quick-turn trailer designed to build internal awareness around the company’s AI transformation initiative. JTI needed something that could carry weight in a room full of stakeholders, generate momentum around a new direction, and arrive fast. That combination is exactly where the right AI video workflow earns its place.
Designing the Creative Before Touching a Model
The first decision was format. A traditional corporate explainer video would have felt out of step with the energy JTI wanted to convey. So we leaned into the trailer convention: a fast-paced montage of future story moments, edited with the rhythm and tonal cues audiences already associate with cinematic storytelling.
That decision did real work. It gave us permission to compress a vision into 45 seconds without losing emotional payoff, and it gave the AI generations a coherent stylistic frame to live inside. Trailers thrive on suggestion rather than explanation, which happens to be a sweet spot for generative video.
Treating Every Element as a Generative Ingredient
Once the creative was locked, we moved into design. Characters, environments, and tools were developed as generative ingredients: approved visual references that downstream generations could pull from. This is where production discipline separates strong AI work from generic output.
If you let a model invent a protagonist from scratch each time, you get visual drift across shots. Continuity collapses. By pre-defining the ingredients and feeding them into our node-based workflow as locked inputs, we maintained character likeness, environmental consistency, and tonal coherence across the entire montage.
The Node-Based Workflow
Our generation pipeline is built around a node-based architecture, which is a meaningful detail worth pausing on. Each scene flows through a chain of nodes that govern inputs, prompts, references, and conditioning. Approved ingredients enter at the top. Generated shots come out the bottom. Creative direction is enforced at every node along the way.
The practical effect is control. We can revise a single element without regenerating an entire sequence. We can swap an environment while keeping the character consistent. We can route the same approved style through twenty different shots without the chaos that comes from prompting blind. This is the architecture that makes a two-week timeline real instead of theoretical.
Assembly, Finish, and Delivery
Once the generations were approved, we assembled the montage and treated it the way we treat any finished piece: voiceover recorded and synced, music scored to the cuts, color graded for cohesion across AI-generated frames, sound designed to add weight to the visuals, and graphics layered for brand presence.
The full delivery, from brief to broadcast, took under two weeks. Not because we cut steps, but because the architecture compressed them.
Why This Matters for Brands
Plenty of teams can press a button on a model. Far fewer can deliver a brand-quality piece on a corporate timeline, with directional consistency across every frame and a finish that holds up next to traditional production. The difference is judgment, applied through structure. Designing the creative concept. Building the right ingredients. Routing them through a workflow that protects intent. Finishing with the same care any other broadcast piece would receive.
That is what we do at Lemonlight. The JTI trailer is one example, but there are many more coming in this series.
If you’re ready to speak with one of our experts to see if AI video could be the right fit for your next video marketing campaign, schedule a no-pressure call below and we’ll figure it out together.
Social Video
Video Marketing Strategy
Video Production Tips
