Local MiniMax H3 on RTX 5090: 15s 2K video in ~9 minutes via a 3-stage pipeline
Hailuo_AI · x · 2026-09-08
A Japanese developer details a full local pipeline generating 15-second, 2K+ resolution videos with MiniMax H3 on a single RTX 5090 in about 9 minutes total (measured: 362 frames, 85s inference):
- 480p drafting with Alibaba's distilled LoRA (2m55s) to quickly find good takes
- Refine: same latent upscaled to 1344×768 with 4 refine steps (4m40s)
- RTX Video Super Resolution to 2688×1536 (1m25s; 0.1–0.2s per frame, single pass instead of diffusion re-rendering)
The trade-off: VSR is fast but can't fix a bad base frame, and the 2K output still looks slightly soft — quality-wise, non-distilled 25 steps or native 768p would be better. The appeal is a relaxed workflow: cheap 3-minute drafts, upscale only the keepers via saved latents.
More from Multimodal
- Techhalla shares full 2D-to-3D workflow with Opus 5.5 and Kling 4.0 prompts — techhalla · 2026-10-26
- Rebuilding real events as photoreal SF digital twins, filmed cinematically from a phone — bilawalsidhu · 2026-10-26
- How to turn Claude Code into a video editor: a full reel in 19 minutes — PrajwalTomar_ · 2026-10-26
- "Recursive eyes" prompt idea produces mesmerizing spiral of infinitely recurring faces — LudovicCreator · 2026-10-26
- Black Forest Labs open-sources 5.6B FLUX 3 Image Klein, +91 Elo with 38% fewer params — dl_weekly · 2026-10-26
- Claude + HyperFrames can clone any short-video style in minutes — petergyang · 2026-10-26