AI Content Machine
Stylar
Procedural Content Generation
The Ask
Stylar is an early-stage GenAI fashion try-on mobile app. One of their marketing strategies was to launch a User-Generated Content (UGC) campaign, which involved posting a high volume of content featuring people using their product across multiple social media accounts.
I was brought on to overhaul this process and create a programmatic system to generate the content procedurally using FFMPEG and Python.
The Video Workflow
With a single image as an input, I can procedurally composite the product screenshots using a Python FFmpeg script and make an API call to FLUX Klein to perform the GenAI outfit swap. I also tested making an API call to Claude for text variations, but I found that swapping between 12 handwritten templates was more effective.
Designing the right template for the video thumbnails was also critical in making sure the videos looked unique and professional.
Drag and drop a folder of images and we get hundreds of videos with one click!
We can use this tool to quickly test variations for different demographics.
To automate the sourcing of trending celebrity images, I use Claude co-work with Chrome MCP to source images from fashion news websites.
Video Output Templates
Image Carousel Output Templates
