MiniMax H3 video gen config: LoRA and settings breakdown
erioca · reddit · 2026-08-18
The author shared a detailed configuration for generating videos using the MiniMax H3 model:
- Model: minimaxh3hybridfl2varef2vab20
- Video VAE: minimaxh3videovaeint8convrot
- Res/Duration: 16:9, 1.0, 6 clips total (9 sec each), composited in Inshot.
- Turbo LoRA: larryvrh/MiniMax-H3-Turbo-Lora (600ema)
- Tech Details: ComfyKitchen Attention and Spectrum used (SageAttention patch and Mem Eff Node disabled).
- Ref Images: All refs resized between 1000px-500px (Character 1000px, Background 500px), 4 refs total.
- Inference Time: Average Inference Stage 1100 sec.
Related event: MiniMax H3 Video Generation Tested with LoRA Workflows(3 posts)→
More from Multimodal
- Inside 'The Very Next Day': 2D keyframes plus a custom IC-Lora by LTX — paultrillo · 2026-10-16
- fal Worlds: designer unveils brand identity for fal's generative world-models platform — adamho · 2026-10-16
- Slopcannon picks Muse Wonder as default AI director, adds frontier model picker — basedjensen · 2026-10-16
- Developer mass-produces 200 videos in hours using Claude Code + Hyperframes — mcraddock · 2026-10-16
- Creator showcases V2 of AI-generated artwork 'Alien Rebel' — creatoroff · 2026-10-16
- Mirage Tesseract lets Claude and ChatGPT create full HDR videos and motion graphics — LinusEkenstam · 2026-10-16