ComfyUI Adds Day 0 Support for MiniMax Video Model, Slashing VRAM by 66% for RTX 3060
crystal_alpine · reddit · 2026-08-03
ComfyUI announced Day 0 support for the newly open-weight MiniMax H3 video model. The model supports text-to-video, image-to-video, and editing, generating clips up to 2K resolution and 15 seconds with real stereo audio.
To run this massive model on consumer hardware, the team performed deep engineering optimizations: pruning 40% of the modulation weights and replacing them with a lookup table, combined with int8 quantization and custom kernels. This reduces total memory footprint by 66% (from 123.6 GB to 42.5 GB). With dynamic VRAM offloading, users can run the 2K video model locally on a single RTX 3060 GPU.
Related event: MiniMax Video Model Goes Viral with Stunning User Tests(61 posts)→
More from Infra
- GitHub rewrote the 800k-line Copilot runtime in Rust using AI agents, cutting CPU and memory by over 90% — bibryam · 2026-10-16
- NVIDIA quietly delists the entry-level RTX 5050 from its marketplace, following the RTX 5090 — chemist_slime · 2026-10-16
- Grok claims 816B MoE model runs locally on dual M5 Ultra Mac Studios via MLX — grok · 2026-10-16
- Same 1,000 Cloudflare Durable Objects: $4,050/month cut to $22 with one trick — threepointone · 2026-10-16
- NVIDIA ships FP4-quantized Qwen3.8-Flash for local use, trending on Hugging Face — nvidia · 2026-10-16
- Baidu's Virtual Head Attention tames the KV cache, FLOPs, quality triangle in long-context LLMs — baidu · 2026-10-16