diff --git a/FlashVSR/examples/WanVSR/infer_flashvsr_v11_full.py b/FlashVSR/examples/WanVSR/infer_flashvsr_v11_full.py
index 3a4c169..dec5122 100644
--- a/FlashVSR/examples/WanVSR/infer_flashvsr_v11_full.py
+++ b/FlashVSR/examples/WanVSR/infer_flashvsr_v11_full.py
@@ -255,8 +255,18 @@ def init_pipeline_v11(prompt_path,LQ_proj_in_path="./FlashVSR/LQ_proj_in.ckpt",c
         pipe.denoising_model().LQ_proj_in.load_state_dict(torch.load(LQ_proj_in_path, map_location="cpu",weights_only=False), strict=True)
 
     pipe.denoising_model().LQ_proj_in.to('cuda')
-    pipe.vae.model.encoder = None
-    pipe.vae.model.conv1 = None
+    if pipe.vae is  None:  # safetensors cause error or unknown safetensors or pth
+        from diffusers import AutoencoderKLWan
+        config=AutoencoderKLWan.load_config(os.path.join(cur_dir,"FlashVSR/config.json"))
+        vae=AutoencoderKLWan.from_config(config).to(device,dtype=torch.bfloat16)
+        vae_dict=torch.load(vae_path, map_location="cpu",weights_only=False) if not vae_path.endswith(".safetensors") else load_file(vae_path,device="cpu")
+        vae.load_state_dict(vae_dict,strict=False)
+        del vae_dict
+        pipe.vae=vae
+        pipe.vae.encoder = None
+    else:
+        pipe.vae.model.encoder = None
+        pipe.vae.model.conv1 = None
     #pipe.to('cuda'); 
     pipe.enable_vram_management(num_persistent_param_in_dit=None)
     pipe.init_cross_kv(prompt_path); pipe.load_models_to_device(["dit","vae"])