Realistic Talking Avatar Videos

I have a reference video of a real presenter, a high-resolution photo, and a finished voice-over track in MP3. Your job is to combine them into a deep-fake style talking-head—think HeyGen level realism—that lip-syncs perfectly to the audio. The videos are for internal training, so natural mouth movement, eye-blinks, and subtle facial expressions are essential.

Workflow
1. Proof of concept: create a short 30–60-second sample. I supply all assets; you return a 720p render with a transparent background (alpha channel) so I can drop it onto any slide or scene.
2. Full production: once the test is approved, convert 5–8 complete sessions, each 15–20 minutes long, to the same 720p, transparent-background spec. H.264 or ProRes files are fine; please also send the project files for archiving.

What I need to see from you
• A portfolio (or secure links) that shows realistic face-replacement or avatar work—preferably training or educational content rather than short gimmicks.
• Your pipeline (e.g., DeepFaceLab, FaceFusion, Stable Diffusion + SADTalker, etc.) and typical turnaround per finished minute.
• Any constraints on input formats or lighting you need to achieve the most believable result.

Once we confirm the approach I can upload the test assets immediately and we can get started.