🎁 Fan Registration Bonus
Register to receive 1000 RH coins:
🔗 https://www.runninghub.ai/?inviteCode=rh-v1325
⚙️ Workflow
Ultimate Simplified MiniMax H3 Full-Feature Auto Prompt Workflow (Lenient moderation, NSFW content allowed)
🔗 Experience link: https://www.runninghub.ai/post/2088114005419282433/?inviteCode=rh-v1325
📦 Models
🔗 https://civarchive.com/models/2879272/minimax-h3-remix
⚡ Turbo LoRA
🔗 https://civarchive.com/api/download/models/3257181?fileId=3140963
🎬 Clip (uncensored）
🔗 https://civarchive.com/api/download/models/3254025?fileId=3137729
🛠️ Required Nodes
- Easy-H3: https://github.com/FX-FeiHou/ComfyUI-FeiHou-Easy-H3
- FeiHou-Toolbox: https://github.com/FX-FeiHou/ComfyUI-FeiHou-Toolbox
✅ This FeiHou Easy H3 Remix workflow is custom‑built for the MiniMax H3 Remix V0.6 test‑version model. Adopting an integrated architecture, it delivers a complete set of creative capabilities to build an efficient AI‑video generation pipeline for video creation, digital‑human production, image secondary creation and other scenarios.
🖼️ The workflow supports infinite‑canvas‑style operation. It accepts multi‑modal reference materials including images, videos and audio. Reference resources can be flexibly added, removed and swapped without frequent node rewiring, enabling free arrangement of generation logic and streamlining the overall creation process.
🔧 Built upon hierarchical‑recombination fusion technology, it combines capabilities from Fl2VA and REF2VA models. Lightweight LoRAs for limb and motion optimization are embedded to mitigate artifacts such as distorted limbs and stiff movements while preserving base‑model stability, enhancing dynamic performance and visual integrity. Matched exclusive uncensored CLIP model improves prompt compatibility and expands creative possibilities.
💻 Two precision options are provided for diverse hardware environments. The BF16 full‑precision variant prioritizes fine‑detail reconstruction for high‑quality output. The INT8 ConvRot low‑VRAM variant is optimized for limited‑hardware setups, reducing memory footprint while maintaining output quality and operational stability.
📝 Built‑in prompt inversion and intelligent expansion modules extract visual descriptions from reference inputs and enrich prompt texts, easing prompt‑engineering difficulties and improving alignment between prompts and generated outputs.
⚙️ A dual‑sampling generation architecture is implemented. The first‑stage sampling produces base composition, while the second‑stage sampling refines fine details. It optimizes textures, motions and color rendering while preserving original composition and subject identity. The second‑stage sampling can be toggled independently to balance generation speed and visual quality.
📦 Pre‑wired nodes and preset parameters allow out‑of‑the‑box usage. It lowers the barrier to setup and tuning for MiniMax H3 and satisfies various multi‑modal video‑generation requirements.