This is a partial NVFP4 quantization of <a target="_blank" rel="ugc" href="https://huggingface.co/lightx2v/Wan2.1-I2V-14B-480P-StepDistill-CfgDistill-Lightx2v">Wan2.1-I2V-14B-480P-StepDistill-CfgDistill-Lightx2v</a> by lightx2v, produced using <a target="_blank" rel="ugc" href="https://github.com/silveroxides/convert_to_quant">convert_to_quant</a> by <a target="_blank" rel="ugc" href="https://huggingface.co/silveroxides">silveroxides</a>.<a target="_blank" rel="ugc" href="https://huggingface.co/lightx2v/Wan2.1-I2V-14B-480P-StepDistill-CfgDistill-Lightx2v">Wan2.1-I2V-14B-480P-StepDistill-CfgDistill-Lightx2v</a> is an image-to-video generation model built on <a target="_blank" rel="ugc" href="https://huggingface.co/Wan-AI/Wan2.1-I2V-14B-480P">Wan2.1-I2V-14B-480P</a>. It applies step distillation and classifier-free guidance distillation to reduce inference to 4 steps without CFG, cutting generation time substantially while preserving output quality. IMPORTANTSince NVFP4 is only supported on NVIDIA Blackwell architecture GPUs, running this model requires a Blackwell GPU with its corresponding support enabled in torch, along with a recent version of ComfyUI and <a target="_blank" rel="ugc" href="https://github.com/Comfy-Org/comfy-kitchen">comfy-kitchen</a> built against CUDA 13.The model can be used in ComfyUI with the following parameters, based on the distilled model's own recommendations:<ul><li>Shift: 5.0</li><li>Sampler: LCM</li><li>Scheduler normal</li><li>CFG: 1.0</li><li>Steps: 4</li></ul>Other combinations are possible, such as euler/simple and heun/linear_quadratic.

Wan2.1 I2V Lightx2v Step/CFG NVFP4

This version tries to protect the following layers, among others. <ul><li>cross_attn.k</li><li>cross_attn.q</li><li>self_attn.k</li><li>self_attn.q</li></ul>

wan21I2VLightx2vStep_v10.safetensors

wan_2.1_vae.safetensors

v1.0

<h2 id="wan-video">Wan Video</h2>Note: There are other Wan Video files hosted on Civitai - these may be duplicates, but this model card is primarily to host the files used by Wan Video in the Civitai Generator. 👍 SOTA Performance: Wan2.1 consistently outperforms existing open-source models and state-of-the-art commercial solutions across multiple benchmarks.<ul><li>👍 Supports Consumer-grade GPUs: The T2V-1.3B model requires only 8.19 GB VRAM, making it compatible with almost all consumer-grade GPUs. It can generate a 5-second 480P video on an RTX 4090 in about 4 minutes (without optimization techniques like quantization). Its performance is even comparable to some closed-source models.</li><li>👍 Multiple Tasks: Wan2.1 excels in Text-to-Video, Image-to-Video, Video Editing, Text-to-Image, and Video-to-Audio, advancing the field of video generation.</li><li>👍 Visual Text Generation: Wan2.1 is the first video model capable of generating both Chinese and English text, featuring robust text generation that enhances its practical applications.</li><li>👍 Powerful Video VAE: Wan-VAE delivers exceptional efficiency and performance, encoding and decoding 1080P videos of any length while preserving temporal information, making it an ideal foundation for video and image generation.</li></ul> Original HuggingFace Repo: <a target="_blank" rel="ugc" href="https://huggingface.co/Wan-AI">https://huggingface.co/Wan-AI</a>

Files with SHA256: 64a7ef761bfccbadbaa3da77366aac4185a6c58fa5de5f589b42a65bcc21f161