Summary
According to a post on X by @michaelaubry, feeding a video model the raw clip let the original face leak through the swap, and his result came back "as Leo with my beard." The fix he describes is to hide the person first using mask, depth and Face Mesh inputs, then run the face swap, then lay the real audio back on. Each step is organized as one Wireflow board. The post had about 7 likes and 340 views when captured, and a separate tip on syncing cuts to music beats was also mentioned.
AI-assisted summary based on the listed source.
What happened
michaelaubry found that handing a video model the raw clip lets the original face leak through (his came back "as Leo with my beard"), and fixes it by hiding the person first: mask, depth and Face Mesh, then the swap, then the real audio laid back on, each step one Wireflow board (about 7 likes, 340 views)....
Why it matters
Creators using AI face swaps on video would be affected if this leak problem is common, since the fix adds masking, depth and audio steps to the workflow. The post is one creator's account with low engagement, and the snippet shows no independent testing. Watch for other creators reproducing these steps before treating the method as reliable.
Signal Intelligence
Signal Strength 81%
Technical label WATCH
Public Interest 30
Category CREATOR TOOLS
Reader Depth TECHNICAL
Signal Strength reflects source quality, relevance, freshness and evidence. Public Interest helps organize discovery; it is not proof of truth.
Public Interest components
Recognizable Entity Score 0
Practical Impact Score 8
Novelty Interest Score 94
Consequence Score 10
Curiosity Score 32
Shareability Score 46
Why this is here
VQV surfaced this signal because it is recent, relevant to AI Video, connected to Scobleizer News Pulse (X).
