Building the full creative pipeline for Meta ad campaigns. From concept and character development through to weekly performance review and iteration.
Video Marketing Manager, CommentSold / POP.STORE
I built and ran the AI video content pipeline for Meta ad campaigns across Instagram and Facebook. The brief was to create content that converted, not just content that looked good. Every creative decision, from which persona to cast to what hook to open on, was made with a conversion outcome already in mind.
Over 100 ad creatives produced across the campaign. Each one a different script, hook, or character variant. Weekly performance reviews fed directly into the next round of work. A cast of AI-generated characters built from scratch, each designed to reach a different audience segment with a different emotional register.
Audience profiling, platform analysis, competitor content audit. Understanding what is already working in the space before deciding what to make.
Visual references, colour palette, aesthetic direction, tone of voice. A clear creative vision documented before a single prompt is written.
Translating the moodboard into precise AI prompts. Character consistency, scene composition, motion direction, output format tailored to the campaign goal.
AI video generation using Kling, Runway ML, Hume AI, and Nanobanana. Iteration, refinement, and quality checking against the brief.
Final output reviewed against campaign objectives. Is the hook strong enough? Does the visual match the audience? Is the call to action clear?
"The data decides what gets made next." Hook rate, hold rate, and CTR scores drive the weekly call: scale, monitor, or kill. That output shapes the next production cycle.
Each character was built through iterative prompt engineering until their visual identity was consistent enough to run in live campaigns. Physicality, wardrobe direction, voice, energy, scene context. Click any card to learn more.
Hook rate. Hold rate. CVR. Every ad, scored.
I built an internal performance dashboard that tracks every ad running in the library, measuring hook rate, hold rate, and CTR, and scores each one with a clear next action: Scale, Monitor, or Fix. It replaced manual guesswork with a real decision system. Here are some of the ads in performance.
Meta campaigns running across Instagram and Facebook simultaneously. Every creative decision, from initial concept to platform-specific adaptation, ran through a single pipeline with weekly performance reviews feeding directly into the next round of work.
The hook stops the scroll. The character builds enough trust in three seconds to earn a click. The message creates enough intent that the landing page has something real to work with. Every creative decision was made with the end goal already in view.