diff --git a/comfy/sd.py b/comfy/sd.py
index a16f2d14f..7f96a791e 100644
--- a/comfy/sd.py
+++ b/comfy/sd.py
@@ -538,7 +538,7 @@ class VAE:
                     self.working_dtypes = [torch.bfloat16, torch.float16, torch.float32]
                     self.memory_used_encode = lambda shape, dtype: 3300 * shape[3] * shape[4] * model_management.dtype_size(dtype)
                     self.memory_used_decode = lambda shape, dtype: 8000 * shape[3] * shape[4] * (16 * 16) * model_management.dtype_size(dtype)
-                else:  # Wan 2.1 VAE
+                else:  # Wan 2.1 VAE 
                     dim = sd["decoder.head.0.gamma"].shape[0]
                     self.upscale_ratio = (lambda a: max(0, a * 4 - 3), 8, 8)
                     self.upscale_index_formula = (4, 8, 8)
@@ -549,8 +549,10 @@ class VAE:
                     ddconfig = {"dim": dim, "z_dim": self.latent_channels, "dim_mult": [1, 2, 4, 4], "num_res_blocks": 2, "attn_scales": [], "temperal_downsample": [False, True, True], "dropout": 0.0}
                     self.first_stage_model = comfy.ldm.wan.vae.WanVAE(**ddconfig)
                     self.working_dtypes = [torch.bfloat16, torch.float16, torch.float32]
-                    self.memory_used_encode = lambda shape, dtype: 6000 * shape[3] * shape[4] * model_management.dtype_size(dtype)
-                    self.memory_used_decode = lambda shape, dtype: 7000 * shape[3] * shape[4] * (8 * 8) * model_management.dtype_size(dtype)
+                    self.memory_used_encode = lambda shape, dtype: (1500 if shape[2]<=4 else 6000) * shape[3] * shape[4] * model_management.dtype_size(dtype)
+                    self.memory_used_decode = lambda shape, dtype: (2200 if shape[2]<=4 else 7000) * shape[3] * shape[4] * (8*8) * model_management.dtype_size(dtype)
+
+
             # Hunyuan 3d v2 2.0 & 2.1
             elif "geo_decoder.cross_attn_decoder.ln_1.bias" in sd: