模型介绍
Hyper-SD是一种用于高效图像合成的轨迹分段一致性模型，能够通过1到8个推理步骤在SDXL和SD1.5上实现最先进的性能。在1步推理中，Hyper-SDXL的CLIP分数超过SDXL-Lightning +0.68，Aes分数超过+0.51。
作者：ByteDance
预览图
模型列表
Hyper-FLUX.1-dev-8steps-lora_rank1.safetensors
Hyper-FLUX.1-dev-16steps-lora.safetensors
Hyper-FLUX.1-dev-8steps-lora.safetensors
Hyper-SD3-16steps-CFG-lora.safetensors
Hyper-SD3-4steps-CFG-lora.safetensors
Hyper-SD3-8steps-CFG-lora.safetensors
Hyper-SD15-1step-lora.safetensors
Hyper-SD15-2steps-lora.safetensors
Hyper-SD15-4steps-lora.safetensors
Hyper-SD15-8steps-CFG-lora.safetensors
Hyper-SD15-8steps-lora.safetensors
Hyper-SD15-12steps-CFG-lora.safetensors
Hyper-SDXL-1step-lora.safetensors
Hyper-SDXL-1step-Unet.safetensors
Hyper-SDXL-1step-Unet-Comfyui.fp16.safetensors
Hyper-SDXL-2steps-lora.safetensors
Hyper-SDXL-4steps-lora.safetensors
Hyper-SDXL-8steps-CFG-lora.safetensors
Hyper-SDXL-8steps-lora.safetensors
Hyper-SDXL-12steps-CFG-lora.safetensors
