Install or open the ChatGPT Desktop App. In the model picker choose Astra (GPT-6).
Do not run this in a thin browser tab if you need the Apps & Connectors panel, Desktop is where Higgsfield attaches. Leave one empty chat ready before you paste anything.
Use the performer-swap prompt word for word. The ready-made prompt is in the members pack below.
It maps left performer → @Image 1 and right performer → @Image 2, and locks audio, camera, cuts, set, and movement. Do not rewrite the lock lines.
Paste once into the chat composer and stop.
Attach the source clip first (any short two-person performance works, hotel lobby, mic stand, interview). Then attach two clear face photos: Image 1 = person who should land on the left, Image 2 = person on the right.
Front-facing, well lit, matching the outfits you want kept. Wrong order swaps the wrong faces.
In ChatGPT Desktop go to Settings → Apps & Connectors. Add Higgsfield, paste your API key, confirm Connected.
In the model / tool row pick Higgsfield Genjutsu. If Genjutsu is missing, the connector is not live, reconnect the key before you hit run.
Send the message. When Genjutsu finishes, watch once with sound on.
Pass only if faces and outfits match the two photos AND every move, cut, and the original audio stay intact. Fail if the camera path drifts or a face melts.
Export the keeper and post. That is the 30-minute loop, one clip, two photos, one run.
This guide is the full process. The ready-made version lives inside AI Creative OS on Skool, next to every other prompt pack I post.
