Claymation body horror on 8GB: three wrong settings, not the GPU, made local renders slow
roro9- · reddit · 2026-10-30
On an 8GB RTX 5060 laptop (WSL2), the author ran a fully local claymation body-horror pipeline: Flux for stills, Wan 2.2 for motion at 12fps. Key finding — slowness was all self-inflicted:
- Quant choice: Flux GGUF Q4KS (6.4GB) offloads 1127MB to CPU at 11-22s/step; the larger FP4 (svdq-fp4r32 with nunchaku, 6.6GB) fits fully in VRAM at 3.4-3.8s/step.
- cpuoffload on auto left the GPU at 1% utilization (234s/step); disabling it fixed that.
- Self-imposed resolution cap: Wan ti2v-5b q5 runs fine at 544x960, 81 frames, 20 steps on 8GB (14-17s/step) — the author had rendered at 480x608 for weeks assuming more wasn't possible.
The skin-peeling effect is actually 7 aligned Flux stills revealed top-down with a mask; Wan only handles the final scream. Every slow render had an idle GPU and zero errors.
More from Infra
- effect-harness on Cloudflare: run agent loops, tools and credentials on separate VMs — samgoodwin89 · 2026-10-31
- Google's Project Suncatcher boots TPU in orbit and beams back first AI response — sundarpichai · 2026-10-31
- Together AI launches Post-Training platform for agentic RL with tool-use environments — togethercompute · 2026-10-31
- General Compute raises $100M for inference capacity as CoreWeave stock hits 12-month low — YvesMulkers · 2026-10-31
- Analyst warns of unstable AI compute feedback loop as Kioxia's NAND price warning triggers flash crash — bookwormengr · 2026-10-31
- SOMA launches enterprise prompt-compression layer claiming 20%+ inference savings — markjeffrey · 2026-10-31