Running Minimax Video on 12GB VRAM: 22-Min Render Tested
Svan_Derh · reddit · 2026-08-06
The author shares hands-on experience running the Minimax video generation model locally on a consumer GPU with 12GB of VRAM.
After tweaking prompts and overcoming Out-of-Memory (OOM) errors, the video was successfully rendered in 22 minutes, plus an additional 5 minutes for upscaling. This demonstrates that with proper parameter tuning, current open-source video models can run on limited hardware.
More from Infra
- 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
- DuoDiscover agents auto-build fast eval proxies, scoring 2.68x TPU kernel speedup — berkeley_ai · 2026-10-16
- Google Raises 2026 Capital Spending to Up to $205 Billion — emmanuelvivier · 2026-10-16