5-second video in 8 minutes on M4 Max: fastest Minimax H3 Mac workflow
xNightWardenx · reddit · 2026-08-26
The author runs Minimax H3 (ref2va) on an M4 Max with 48GB RAM, generating a 480p/24fps/5s video in 20 steps in 7:57 — possibly the fastest Mac workflow right now.
Key components:
- ComfyUI-AppleSilicon-FP8 is required to run H3 in Comfy Desktop on Mac at all
- Official pruned int8 diffusion model + NVFP4/AWQ Qwen3VL-32B text encoder
- Most of the speed comes from a community Turbo LoRA: 8 steps (author uses 20), euler sampler, beta scheduler, LoRA strength 1.0
- The Spectrum node saves another 30% generation time
- A live preview node (Kijai's TAE + KJNodes) costs 20s but lets you kill bad renders early
48GB RAM pitfall: loading the Turbo LoRA causes a memory spike that can crash renders; routing through the official Ref2v workflow's switch (with the if/else model node set to true) works around it.
More from Multimodal
- Gemini Launches Interactive 3D Visualizations Directly in Chat — jocarrasqueira · 2026-08-26
- Single-Word Prompt Challenge: Using 'Heofoncandel' in Midjourney — tisch_eins · 2026-08-26
- Short film 'Grendel' teaser: Night Comes To Heorot — Puzzleheaded_Bet3241 · 2026-08-26
- AVERNUS-9: Space Horror Short Film Clip — Tadeo111 · 2026-08-26
- Wizstar's two-stage pipeline fixes lip-sync and stability issues in AI avatars — Med1_Ai · 2026-08-26
- Paper Demo-ICL Accepted to EMNLP 2026 — liuziwei7 · 2026-08-26