Changing the file metadata to resemble an iPhone recording is unlikely to solve this reliably. Platforms can evaluate the actual visual and audio content, not just the camera information stored in the file. Since you're generating videos with Wan through Higgsfield, I would review Instagram's current AI-content disclosure guidelines and use the appropriate labeling options. If a genuine recording was incorrectly labeled, the relevant support or appeal process is a better route than trying to disguise its origin.
