Local V2V outfit transfer on a 16GB M3 MacBook Air: 15.7 min cut to 2.4 min
might_thinker · reddit · 2026-11-16
A developer got local video-to-video outfit transfer running on a fanless 16GB M3 MacBook Air, cutting processing of a 46-frame (3s) clip from 15.7 minutes to 2.4 minutes, fully offline and with visibly better results. Key optimizations:
- Crop the subject: process only a tight 288×512 region around the outfit and composite back onto the original background;
- Precompute prompts: cached outfit prompts drop prompt loading to 0.05s;
- Overlap workloads: face processing on CPU/Neural Engine while the GPU handles the next diffusion chunk, saving 15s;
- Quick previews: a 5-frame test pass (20s) catches mask issues before a full render.
Remaining limits: 11% performance degradation on the second chunk under sustained load, overly smooth fabric detail, lost detail on distant subjects, and a pipeline that isn't plug-and-play. The takeaway: treating unified memory and compute scheduling as constraints beats throwing a bigger model at the problem.
More from Infra
- BIS issues rule blocking remote access to US AI chips via third-country data centers — teortaxesTex · 2026-11-16
- Qdrant 1.20.1: 50% fewer cold-storage reads, ~2x faster quantization, agent-ready setup — qdrant_engine · 2026-11-16
- One 16GB GPU runs songs, voice, images, 720p video and local RAG — all offline and open source — llama-of-death · 2026-11-16
- Starlink Now Protects Greek Trains From Collisions With 5cm-Accurate Tracking of 134 Locomotives — XFreeze · 2026-11-16
- Google donates Agent Substrate, gVisor and OpenRL to CNCF; Agent Substrate 1.0 goes GA on GKE — dl_weekly · 2026-11-16
- Nvidia scandal isn't slowing down: Huang's near-$100B power web under strain — The Verge AI · 2026-11-16