Consistent Person AI Compositing

I have a set of still images (JPEG/PNG) and short MP4 clips that need one specific person added into each scene. You will receive clear reference photos and a few reference frames of that same individual; the task is to place this exact person—not a new AI-generated face—into every supplied file while matching pose, lighting, perspective, and grain so the composites look genuinely shot on location.

The workflow is up to you, but I expect a stable, repeatable identity pipeline: face-tracking or key-point extraction for the videos, mask refinement, colour matching, and final grading so the added person blends perfectly frame to frame and shot to shot. Tools such as Photoshop, After Effects, Runway, Stable Diffusion ControlNet or similar are all fine as long as the end result is seamless.

Deliverables
• Final retouched JPEG/PNG images with the person integrated
• Final graded MP4 files with the person integrated, same duration and resolution as originals
• Source project files (PSD, AEP, or relevant node/timeline files) so I can make minor tweaks later

Acceptance criteria
• The same face and features appear consistently across every asset
• No visible halos, mismatched lighting, or jitter in motion scenes
• All files maintain the original resolution, codec, and aspect ratio

If you have prior examples of identity-consistent compositing, feel free to share a link so I can see the quality I can expect.