How I built an end-to-end AI face swap video workflow, from shot planning to frame alignment
Few-Profession421 · reddit · 2026-10-21
A complete breakdown of an AI face-swap video pipeline organized into nine folders (raw source → shot planning → scene references → prompt packages → generations → frame alignment → final assembly → QC → archive).
Key practices:
- Use ffprobe to record frame count/rate and SHA256 checksums to catch wrong exports
- Cut shots at natural motion boundaries, not arbitrary timestamps; for multi-person scenes, explicitly specify clothing, left/right positioning, and occlusion
- Generate reference stills before video: at least start/middle/end frames per shot, checked for identity, hairstyle, clothing, pose, hands, and layering — if the still is wrong, video generation won't fix it
- Package every shot as a reproducible task: stills + 24 FPS motion-reference video + motion description + ComfyUI config + checksum manifest
- Two-pass resolutions (1280×736 → 1920×1088); motion prompts describe only visible actions
- Pitfall: a Musk replacement character closed his eyes too early — references reduce errors but can't guarantee correct motion
More from coding & agent
- Neo4j GraphAcademy launches AI proxy so workshop learners skip OpenAI billing setup — JeremyCMorgan · 2026-10-21
- Three weekend Claude Code projects: personal benchmarks, mini-SaaS, AI asset library — EXM7777 · 2026-10-21
- vLLM gains sleep mode on Apple Silicon: idle model servers now return KV cache memory — abhijithneil · 2026-10-21
- Open-source Diagram Design turns prompts and codebases into editorial-grade technical diagrams — bibryam · 2026-10-21
- Agent web-data startup raises $9M led by General Catalyst, 800+ customers — aftahi_ai · 2026-10-21
- AWS shows how to build conversational supply chain agents with Bedrock AgentCore and MCP — AWS ML Blog · 2026-10-21