<h3 id="this-workflow-contains-system-prompts-of-llm-prompt-enhancer-for-all-t2va-i2va-l2va-fl2va-and-ref2va.">This workflow contains system prompts of LLM prompt enhancer for all T2VA, I2VA, L2VA, FL2VA and Ref2VA.</h3><h3 id="it-supports-local-gguf-llm-models.-e.g.-gemma-4-qwen-3.6">It supports local gguf LLM models. e.g. Gemma 4, Qwen 3.6</h3><h3 id="no-api-needed.">No API needed.</h3><h3 id="purpose:-to-make-user-prompt-conform-with-minimax-official-prompting-guide.">Purpose: To make user prompt conform with MiniMax official prompting guide.</h3><h3></h3><ul><li><p>Choose a big MoE LLM model (i.e. something with -A?B) instead of a small non-MoE model.</p></li><li><p>Set "cpu_moe": true to speed up LLM with limited vram. (Q8 Gemma 4 26B-A4B model needs only 6GB vram.)</p></li><li><p>To set image_min_tokens for Gemma 4, follow n_ubatch &gt; image_max_tokens &gt; image_min_tokens. (e.g. 2240, 2240, 560)</p></li><li><p>Qwen 3.6 (e.g. Q6 35B-A3B) can also be used instead of Gemma 4. Just set image_min_tokens to 1024 and n_ctx to a larger value (e.g. 16384).</p></li></ul><p></p><h3 id="custom-node-used:">Custom node used:</h3><h3 id="comfyui_simple_qwen3-vl-gguf"><a target="_blank" rel="ugc" href="https://github.com/KLL535/ComfyUI_Simple_Qwen3-VL-gguf">ComfyUI_Simple_Qwen3-VL-gguf</a></h3><p>ComfyUI_Simple_Qwen3-VL-gguf requires installation of llama-cpp-python wheel and Nvidia CUDA Toolkit. (Comfyui's built-in CUDA might not work.)</p><p>e.g.</p><p>cmd</p><p>cd /d C:\StabilityMatrix\Data\Packages\ComfyUI\venv\Scripts\</p><p>python.exe -m pip install ???.whl</p><h3 id="comfyui-solattn_triton"><a target="_blank" rel="ugc" href="https://github.com/kijai/ComfyUI-SolAttn_triton">ComfyUI-SolAttn_triton</a></h3><h3></h3><h3 id="system-prompts-modified-from:">System prompts modified from:</h3><h3 id="h3_llm_instructions"><a target="_blank" rel="ugc" href="https://github.com/cushycrux/H3_LLM_Instructions">H3_LLM_Instructions</a></h3><h3 id="rzgar"><a target="_blank" rel="ugc" href="https://huggingface.co/MiniMaxAI/MiniMax-H3/discussions/28">rzgar</a></h3><h3 id="thanks-for-their-work.-cheers.">Thanks for their work. Cheers.</h3><p></p>

Files with SHA256: 7cfdd92238c216b85205e5bdeaf1c7a45c06c8556047d5151dd31a19923ec7f3