diff --git a/README.md b/README.md
index 968716e..1ae7280 100644
--- a/README.md
+++ b/README.md
@@ -37,11 +37,11 @@ Loads the OmniVoice model. Downloads automatically from HuggingFace on first run
 
 | Input | Type | Description |
 |-------|------|-------------|
-| `model_source` | dropdown | `Auto-download (HuggingFace)` or `Local path` |
-| `local_path` | string | Path to local checkpoint (optional) |
 | `device` | dropdown | `cuda:0`, `cuda:1`, or `cpu` |
 | `dtype` | dropdown | `float16`, `bfloat16`, or `float32` |
 
+Downloads automatically from HuggingFace on first run and caches to `ComfyUI/models/omnivoice/`.
+
 **Output:** `OMNIVOICE_MODEL`
 
 ---
diff --git a/nodes/loader.py b/nodes/loader.py
index f1d1ee7..80fefe5 100644
--- a/nodes/loader.py
+++ b/nodes/loader.py
@@ -25,10 +25,6 @@ class OmniVoiceModelLoader:
     def INPUT_TYPES(cls):
         return {
             "required": {
-                "model_source": (
-                    ["Auto-download (HuggingFace)", "Local path"],
-                    {"default": "Auto-download (HuggingFace)"},
-                ),
                 "device": (
                     ["cuda:0", "cuda:1", "cpu"],
                     {"default": "cuda:0"},
@@ -38,9 +34,6 @@ class OmniVoiceModelLoader:
                     {"default": "float16"},
                 ),
             },
-            "optional": {
-                "local_path": ("STRING", {"default": ""}),
-            },
         }
 
     RETURN_TYPES = ("OMNIVOICE_MODEL",)
@@ -48,19 +41,14 @@ class OmniVoiceModelLoader:
     FUNCTION = "load_model"
     CATEGORY = "OmniVoice"
 
-    def load_model(self, model_source, device, dtype, local_path=""):
+    def load_model(self, device, dtype):
         if OmniVoice is None:
             raise ImportError(
-                "omnivoice is not installed. Run: pip install omnivoice"
+                "omnivoice is not installed. Run: pip install omnivoice --no-deps"
             )
 
-        if model_source == "Local path" and local_path:
-            source = local_path
-        else:
-            source = "k2-fsa/OmniVoice"
-
         model = OmniVoice.from_pretrained(
-            source,
+            "k2-fsa/OmniVoice",
             device_map=device,
             dtype=DTYPE_MAP[dtype],
             cache_dir=CACHE_DIR,
diff --git a/tests/test_loader.py b/tests/test_loader.py
index aa8d8a1..5bf31d9 100644
--- a/tests/test_loader.py
+++ b/tests/test_loader.py
@@ -1,6 +1,5 @@
 # tests/test_loader.py
 from unittest.mock import patch, MagicMock
-import torch
 import pytest
 from nodes.loader import OmniVoiceModelLoader
 
@@ -8,52 +7,38 @@ from nodes.loader import OmniVoiceModelLoader
 def test_input_types_structure():
     inputs = OmniVoiceModelLoader.INPUT_TYPES()
     required = inputs["required"]
-    assert "model_source" in required
     assert "device" in required
     assert "dtype" in required
-    optional = inputs.get("optional", {})
-    assert "local_path" in optional
+    assert "optional" not in inputs or "local_path" not in inputs.get("optional", {})
 
 
-def test_input_types_model_source_choices():
+def test_input_types_device_choices():
     inputs = OmniVoiceModelLoader.INPUT_TYPES()
-    choices = inputs["required"]["model_source"][0]
-    assert "Auto-download (HuggingFace)" in choices
-    assert "Local path" in choices
+    choices = inputs["required"]["device"][0]
+    assert "cuda:0" in choices
+    assert "cpu" in choices
 
 
 def test_return_type():
     assert OmniVoiceModelLoader.RETURN_TYPES == ("OMNIVOICE_MODEL",)
 
 
-def test_load_model_auto_download():
+def test_load_model():
     loader = OmniVoiceModelLoader()
     mock_model = MagicMock()
     with patch("nodes.loader.OmniVoice") as MockOmniVoice:
         MockOmniVoice.from_pretrained.return_value = mock_model
-        result = loader.load_model(
-            model_source="Auto-download (HuggingFace)",
-            device="cpu",
-            dtype="float32",
-            local_path="",
-        )
+        result = loader.load_model(device="cpu", dtype="float32")
     assert result == (mock_model,)
-    MockOmniVoice.from_pretrained.assert_called_once()
-    call_kwargs = MockOmniVoice.from_pretrained.call_args
-    assert call_kwargs[0][0] == "k2-fsa/OmniVoice"
+    call_args = MockOmniVoice.from_pretrained.call_args
+    assert call_args[0][0] == "k2-fsa/OmniVoice"
 
 
-def test_load_model_local_path():
+def test_load_model_dtype_mapped():
+    import torch
     loader = OmniVoiceModelLoader()
-    mock_model = MagicMock()
     with patch("nodes.loader.OmniVoice") as MockOmniVoice:
-        MockOmniVoice.from_pretrained.return_value = mock_model
-        result = loader.load_model(
-            model_source="Local path",
-            device="cpu",
-            dtype="float32",
-            local_path="/some/local/path",
-        )
-    assert result == (mock_model,)
-    call_args = MockOmniVoice.from_pretrained.call_args[0][0]
-    assert call_args == "/some/local/path"
+        MockOmniVoice.from_pretrained.return_value = MagicMock()
+        loader.load_model(device="cpu", dtype="float16")
+    call_kwargs = MockOmniVoice.from_pretrained.call_args[1]
+    assert call_kwargs["dtype"] == torch.float16
diff --git a/workflows/voice_cloning.json b/workflows/voice_cloning.json
index 87b5348..2b7b6cb 100644
--- a/workflows/voice_cloning.json
+++ b/workflows/voice_cloning.json
@@ -20,7 +20,7 @@
         }
       ],
       "properties": {"Node name for S&R": "OmniVoiceModelLoader"},
-      "widgets_values": ["Auto-download (HuggingFace)", "cuda:0", "float16", ""]
+      "widgets_values": ["cuda:0", "float16"]
     },
     {
       "id": 2,