This lora aims to provide a candid social media vibe to minimax image and video generations. It should improve : Skin detail, Composition, lighting, shadows and more! Works with T2V & REF2V I have released the early free version for everybody on Civitai and the further trained checkpoints (4350 steps +) will be available in the sub section for a more pronounced effect. Image samples attached. Enjoy! https://civitai.red/models/2924293/icy-realism-h3?modelVersionId=3308945

Going over my version of the Minimax H3 Reference workflow its messy in the vid but actual workflow is cleaned up. Everything should be in the attachments to this post. Ref model: https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/diffusion_models/minimax_h3_ref2va_pruned_int8_convrot.safetensors Text encoder:https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/text_encoders/qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors Vaes: https://huggingface.co/Comfy-Org/MiniMax-H3/tree/main/vae 4steps lora: https://huggingface.co/lightx2v/Minimax-h3-Turbo/resolve/main/minimax_h3_fl2v_turbo_4step_v1.0_768p_comfyui_bf16.safetensors 8steps lora: https://huggingface.co/lightx2v/Minimax-h3-Turbo/resolve/main/minimax_h3_fl2v_turbo_8step_v1.0_comfyui_bf16.safetensors Most custom nodes will we available thru comfyui Manager Icy megapixel resize custom node included just unzip in custom_nodes and prompting guide included, feed it to a LLM as system prompt Sub Klein based character sheet workflow: https://www.skool.com/ai-tech-and-comfyui-workflows-2226/classroom/bc89a535?md=1acd0b35ad4649619e349a0b034bc0b2

How it works: Plug in images references> LLM creates a prompt based on those > Generate an image using all those references The most important part in this workflow is the prompting, in my case, the LLM I use for the prompting is running on my second computer on the same network which means I don't need to share VRAM between Comfy and LM studio. This will not be the case for most people so you should check unload_after in my LM studio node if you don't have a lot of VRAM to spare. You need to install LM studio and download the LLM you want to use for prompting. I might gift my prompting node if you ask Main model: https://huggingface.co/smhfacct/Minimax-H3-fl2va-ref2va-hybrid-models/resolve/main/minimax_h3_hybrid_fl2va_ref2va_b25-49-int8.safetensors Clip: https://huggingface.co/Comfy-Org/MiniMax-H3/blob/main/text_encoders/qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors 8steps lora: https://huggingface.co/lightx2v/Minimax-h3-Turbo/resolve/main/minimax_h3_fl2v_turbo_8step_v1.0_comfyui_bf16.safetensors Single image VAE:https://huggingface.co/Mamad8/MiniMax-H3-Image-VAE/blob/main/minimax_h3_t1_image_vae_step1597.safetensors Audio VAE (Not used but needed): https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_audio_vae_fp32.safetensors Realism lora: https://civitai.red/models/2200329/galaxyace-lora?modelVersionId=3201619

Quick tutorial on how to install LM studio and the LLM of your choice locally🫡

Hey guys - There are a lot of different video workflows in here. Which one do you guys find to produce the best quality output when it comes to motion control/motion sync? Appreciate it!🙌🏻