Summary
Creator Bart Slodyczka walks through a no-code four-step pipeline for generating varied social media ad content using GPT Astra and a video transformation tool called Higsfield Genjutsu. The core idea: take a single source video clip and use AI to produce multiple variations — swapping backgrounds, changing clothing, replacing the product being held, substituting a different character entirely, and even cloning the voice to match.
The tutorial covers Higsfield’s two main modes — motion transfer and object swap — and shows how to upload a source video alongside reference images to drive specific transformations through natural-language prompts inside the ChatGPT desktop app running Astra. Slodyczka demonstrates several outputs: a living-room clip re-set against an AI-generated outdoor scene with consistent lighting and shadows, a space-suit version, and a gender-swapped version with a matching cloned voice. He also shows how to recycle existing ad library footage by regenerating it in entirely new environments — the same scene rendered once in daylight at a circus and once at night in a dark alleyway.
The video targets business owners and content creators looking to scale ad production without re-shooting. Step four introduces an inspiration tool for avoiding content fatigue. Higsfield offers a free tier with reduced resolution via a shared generation queue, making the workflow accessible without upfront cost.
📺 Source: Bart Slodyczka · Published September 09, 2026
🏷️ Format: Tutorial Demo
