diff --git a/matrixgame/vae_variants/matrixgame_vae.py b/matrixgame/vae_variants/matrixgame_vae.py
new file mode 100644
index 0000000..21e9d7b
--- /dev/null
+++ b/matrixgame/vae_variants/matrixgame_vae.py
@@ -0,0 +1,42 @@
+from .wrapper import VAEWrapper
+import os
+import torch
+from pathlib import Path
+from matrixgame.vae_variants.matrixgame_vae_src import AutoencoderKLCausal3D
+
+class MGVVAEWrapper(VAEWrapper):
+    def __init__(self, vae):
+        self.vae = vae
+        self.vae.enable_tiling()
+        self.vae.requires_grad_(False)
+        self.vae.eval()
+
+    def encode(self, x):
+        x = self.vae.encode(x).latent_dist.sample()
+        if hasattr(self.vae.config, "shift_factor") and self.vae.config.shift_factor:
+            x = (x - self.vae.config.shift_factor) * self.vae.config.scaling_factor
+        else:
+            x = x * self.vae.config.scaling_factor
+        return x
+
+    def decode(self, latents):
+        if hasattr(self.vae.config, "shift_factor") and self.vae.config.shift_factor:
+            latents = latents / self.vae.config.scaling_factor + self.vae.config.shift_factor
+        else:
+            latents = latents / self.vae.config.scaling_factor
+        return self.vae.decode(latents).sample
+
+def get_mg_vae_wrapper(model_path, weight_dtype):
+    if model_path.endswith('.json'):
+        model_path = os.splitext(model_path)[0]
+    config = AutoencoderKLCausal3D.load_config(model_path)
+    vae = AutoencoderKLCausal3D.from_config(config)
+    vae_ckpt = Path(model_path) / "pytorch_model.pt"
+    ckpt = torch.load(vae_ckpt)
+    if "state_dict" in ckpt:
+        ckpt = ckpt["state_dict"]
+    if any(k.startswith("vae.") for k in ckpt.keys()):
+        ckpt = {k.replace("vae.", ""): v for k, v in ckpt.items() if k.startswith("vae.")}
+    vae.load_state_dict(ckpt)
+    vae.to(weight_dtype)
+    return MGVVAEWrapper(vae)