Same Face Across GRWM Loading Frames
A GRWM loading reel fails when the 100% frame is a different person. Lock the face first, then change clothes.
Keeping the same face across GRWM loading frames means the person at 0% and the person at 100% share one identity. Outfit, hair styling, and light can change. Bone structure cannot. If you generate the frames, identity drift is the bug people comment on, and it kills the payoff.
Decide the face before the clothes
Pick one reference still and reuse it for every beat. Do not ask for "a similar woman" on the last frame. Similar is how you get a cousin, not a continuation.
Change one thing per beat
Beat two can add a shirt. Beat three can add hair. Beat four can add the jacket. Changing shirt, hair, room, and face together reads as a new cast member.
Check with the sound off
Lay the five frames in a row and look only at the eyes and jaw. If you would not swear it is one person, do not post. A caption cannot repair a swapped face.
A worked example
Reference: one seated host, shoulder-length hair, left ear visible. 25% adds earrings from that same still. 100% adds a coat. The ear, the mole, and the part in the hair are still there. Anything that loses the mole gets regenerated.
Before you post
- One reference still is used for every beat.
- Only clothes or hair styling change, not the face.
- You can recognise the host with the images in a row.
Where Scro fits
Scro locks the host's face while you change wardrobe, which is the whole point of a loading reel made of stills. You approve each still before motion, and later Life photos can use that same locked person.
Lock the host in Scro. Finished samples are on the examples page.
Related: POV Reel Without Showing Your Face.