Trying out a consistent point-of-view shot with MiniMax H3
retained text
Took a few little prompt adjustments here and there to get H3 to respect point-of-view. I found that if you refer to "the viewer" (ie, "she kicks the viewer"), H3 is more predisposed to include an actual second person. But if you refer to "the camera" (ie, "she kicks the camera"), it's more predisposed to keep the desired point-of-view perspective.
document fetched 2026-09-01 00:32 · canonical https://reddit.com/r/StableDiffusion/comments/1w3k9ti/trying_out_a_consistent_pointofview_shot_with