diff --git a/api/onnx_web/convert/diffusion_original.py b/api/onnx_web/convert/diffusion_original.py
index b1ccae69..602e698e 100644
--- a/api/onnx_web/convert/diffusion_original.py
+++ b/api/onnx_web/convert/diffusion_original.py
@@ -1144,7 +1144,8 @@ def extract_checkpoint(
     extract_ema=False,
     train_unfrozen=False,
     is_512=True,
-    config_file=None,
+    config_file: str =None,
+    vae_file: str =None,
 ):
     """
 
@@ -1306,7 +1307,12 @@ def extract_checkpoint(
         # Convert the VAE model.
         logger.info("converting VAE")
         vae_config = create_vae_diffusers_config(original_config, image_size=image_size)
-        converted_vae_checkpoint = convert_ldm_vae_checkpoint(checkpoint, vae_config)
+
+        if vae_file is None:
+            converted_vae_checkpoint = convert_ldm_vae_checkpoint(checkpoint, vae_config)
+        else:
+            vae_checkpoint = safetensors.torch.load_file(vae_file, device="cpu")
+            converted_vae_checkpoint = convert_ldm_vae_checkpoint(vae_checkpoint, vae_config)
 
         vae = AutoencoderKL(**vae_config)
         vae.load_state_dict(converted_vae_checkpoint)
@@ -1427,7 +1433,7 @@ def convert_diffusion_original(
         logger.info("torch pipeline already exists, reusing: %s", torch_path)
     else:
         logger.info("converting original Diffusers check to Torch model: %s -> %s", source, torch_path)
-        extract_checkpoint(ctx, torch_name, source, config_file=model.get("config"))
+        extract_checkpoint(ctx, torch_name, source, config_file=model.get("config"), vae_file=model.get("vae"))
         logger.info("converted original Diffusers checkpoint to Torch model")
 
     convert_diffusion_stable(ctx, model, working_name)