Local Video Gen with MiniMaxH3: Workflow and Hardware Upgrade Notes
Last-Pie8057 · reddit · 2026-08-10
A user shares their first attempt and complete workflow for local video editing using the MiniMaxH3 model.
- Generation Experience: Overall performance is impressive. It only took two generations to get a usable eating scene, outperforming other local models that typically butcher details. Close-up shots were found to yield better results than wide shots.
- Upscaling Solution: After finding SeedVr2 too slow and lacking in quality, the user opted for FlashVSR (8 mins) combined with RTX Superscale to achieve true 2K resolution, before downscaling to 1080p using ClipChamp.
- Hardware Dilemma: Currently running on a 4080 laptop (12GB VRAM) and 64GB RAM. Anticipating heavy future compute needs, the author is considering building an RTX 5090 desktop despite crazy current prices, predicting hardware costs won't drop in the next 3-5 years.
Related event: MiniMax H3 Local Deployment Tested: Runs on 8GB VRAM, High-End Cards Shine(12 posts)→
More from Infra
- Zero's next-gen IVM engine runs in a single DO, live-querying a 100GiB R2 database — threepointone · 2026-10-16
- Meme: driving to the chip factory in 2028 captures AI's compute anxiety — beffjezos · 2026-10-16
- One volunteer maintains the time zone file behind 4 billion phones — danbri · 2026-10-16
- DeepSeek V4.1 Flash on 4x DGX Spark: weights alone are 511GB, the 494 tok/s claim doesn't fit — Aroochacha · 2026-10-16
- Open-sourced Intel GPU Vulkan inference upgrades: 2x ingestion, 60%+ faster inference — Fun-Aardvark-1143 · 2026-10-16
- ELSFP pain points: >3dB insertion loss, costly PMF, harsh packaging requirements — jwt0625 · 2026-10-16