feat: Add Windows setup and run scripts, update README

This commit introduces automation scripts for Windows users and updates the README.md accordingly.

New/Modified Windows Scripts:
- setup_windows.bat: New script to automate Python checks, ffmpeg warning, virtual environment (.venv) creation, pip upgrade, and dependency installation.
- run_windows.bat: New script to run the application with CPU execution provider by default, activating .venv.
- run-cuda.bat: Updated to use .venv and pass arguments.
- run-directml.bat: Updated to use .venv and pass arguments.

README.md Changes:
- Updated the "For Windows:" section under "Installation (Manual)" to detail the new automated setup using setup_windows.bat and the revised run scripts.
- Recommended Python 3.10 for Windows for best compatibility.
- Provided updated manual setup notes for Windows, including a PowerShell command for ffmpeg installation and using .venv for consistency.
- Ensured the general Python recommendation in the manual setup prerequisites also mentions Python 3.10.

README.md

@@ -134,12 +134,57 @@ Place these files in the "**models**" folder.
 We highly recommend using a `venv` to avoid issues.
 
 
-For Windows:
-```bash
-python -m venv venv
-venv\Scripts\activate
-pip install -r requirements.txt
-```
+**For Windows:**
+
+It is highly recommended to use Python 3.10 for Windows for best compatibility with all features and dependencies.
+
+**Automated Setup (Recommended):**
+
+1.  **Run the setup script:**
+    Double-click `setup_windows.bat` or run it from your command prompt:
+    ```batch
+    setup_windows.bat
+    ```
+    This script will:
+    *   Check if Python is in your PATH.
+    *   Warn if `ffmpeg` is not found (see "Manual Steps / Notes" below for ffmpeg help).
+    *   Create a virtual environment named `.venv` (consistent with macOS setup).
+    *   Activate the virtual environment for the script's session.
+    *   Upgrade pip.
+    *   Install Python packages from `requirements.txt`.
+    Wait for the script to complete. It will pause at the end; press any key to close the window if you double-clicked it.
+
+2.  **Run the application:**
+    After setup, use the provided `.bat` scripts to run the application. These scripts automatically activate the correct virtual environment:
+    *   `run_windows.bat`: Runs the application with the CPU execution provider by default. This is a good starting point if you don't have a dedicated GPU or are unsure.
+    *   `run-cuda.bat`: Runs with the CUDA (NVIDIA GPU) execution provider. Requires an NVIDIA GPU and CUDA Toolkit installed (see GPU Acceleration section).
+    *   `run-directml.bat`: Runs with the DirectML (AMD/Intel GPU on Windows) execution provider.
+
+    Example: Double-click `run_windows.bat` to launch the UI, or run from a command prompt:
+    ```batch
+    run_windows.bat --source path\to\your_face.jpg --target path\to\video.mp4
+    ```
+
+**Manual Steps / Notes:**
+
+*   **Python:** Ensure Python 3.10 is installed and added to your system's PATH. You can download it from [python.org](https://www.python.org/downloads/).
+*   **ffmpeg:**
+    *   `ffmpeg` is required for video processing. The `setup_windows.bat` script will warn if it's not found in your PATH.
+    *   An easy way to install `ffmpeg` on Windows is to open PowerShell as Administrator and run:
+        ```powershell
+        Set-ExecutionPolicy Bypass -Scope Process -Force; [System.Net.ServicePointManager]::SecurityProtocol = [System.Net.ServicePointManager]::SecurityProtocol -bor 3072; iex ((New-Object System.Net.WebClient).DownloadString('https://community.chocolatey.org/install.ps1')); choco install ffmpeg -y
+        ```
+        Alternatively, download from [ffmpeg.org](https://ffmpeg.org/download.html), extract the files, and add the `bin` folder (containing `ffmpeg.exe`) to your system's PATH environment variable. The original README also linked to a [YouTube guide](https://www.youtube.com/watch?v=OlNWCpFdVMA) or `iex (irm ffmpeg.tc.ht)` via PowerShell.
+*   **Visual Studio Runtimes:** If you encounter errors during `pip install` for packages that compile C code (e.g., some scientific computing or image processing libraries), you might need the [Visual Studio Build Tools (or Runtimes)](https://visualstudio.microsoft.com/visual-cpp-build-tools/). Ensure "C++ build tools" (or similar workload) are selected during installation.
+*   **Virtual Environment (Manual Alternative):** If you prefer to set up the virtual environment manually instead of using `setup_windows.bat`:
+    ```batch
+    python -m venv .venv 
+    .venv\Scripts\activate.bat
+    python -m pip install --upgrade pip
+    python -m pip install -r requirements.txt
+    ```
+    (The new automated scripts use `.venv` as the folder name for consistency with the macOS setup).
+
 For Linux:
 ```bash
 # Ensure you use the installed Python 3.10
@@ -150,22 +195,64 @@ pip install -r requirements.txt
 
 **For macOS:**
 
-Apple Silicon (M1/M2/M3) requires specific setup:
+For a streamlined setup on macOS, use the provided shell scripts:
 
-```bash
-# Install Python 3.10 (specific version is important)
-brew install python@3.10
+1.  **Make scripts executable:**
+    Open your terminal, navigate to the cloned `Deep-Live-Cam` directory, and run:
+    ```bash
+    chmod +x setup_mac.sh
+    chmod +x run_mac*.sh
+    ```
 
-# Install tkinter package (required for the GUI)
-brew install python-tk@3.10
+2.  **Run the setup script:**
+    This will check for Python 3.9+, ffmpeg, create a virtual environment (`.venv`), and install required Python packages.
+    ```bash
+    ./setup_mac.sh
+    ```
+    If you encounter issues with specific packages during `pip install` (especially for libraries that compile C code, like some image processing libraries), you might need to install system libraries via Homebrew (e.g., `brew install jpeg libtiff ...`) or ensure Xcode Command Line Tools are installed (`xcode-select --install`).
 
-# Create and activate virtual environment with Python 3.10
-python3.10 -m venv venv
-source venv/bin/activate
+3.  **Activate the virtual environment (for manual runs):**
+    After setup, if you want to run commands manually or use developer tools from your terminal session:
+    ```bash
+    source .venv/bin/activate
+    ```
+    (To deactivate, simply type `deactivate` in the terminal.)
 
-# Install dependencies
-pip install -r requirements.txt
-```
+4.  **Run the application:**
+    Use the provided run scripts for convenience. These scripts automatically activate the virtual environment.
+    *   `./run_mac.sh`: Runs the application with the CPU execution provider by default. This is a good starting point.
+    *   `./run_mac_cpu.sh`: Explicitly uses the CPU execution provider.
+    *   `./run_mac_coreml.sh`: Attempts to use the CoreML execution provider for potential hardware acceleration on Apple Silicon and Intel Macs.
+    *   `./run_mac_mps.sh`: Attempts to use the MPS (Metal Performance Shaders) execution provider, primarily for Apple Silicon Macs.
+
+    Example of running with specific source/target arguments:
+    ```bash
+    ./run_mac.sh --source path/to/your_face.jpg --target path/to/video.mp4
+    ```
+    Or, to simply launch the UI:
+    ```bash
+    ./run_mac.sh
+    ```
+
+**Important Notes for macOS GPU Acceleration (CoreML/MPS):**
+
+*   The `setup_mac.sh` script installs packages from `requirements.txt`, which typically includes a general CPU-based version of `onnxruntime`.
+*   For optimal performance on Apple Silicon (M1/M2/M3) or specific GPU acceleration, you might need to install a different `onnxruntime` package *after* running `setup_mac.sh` and while the virtual environment (`.venv`) is active.
+*   **Example for `onnxruntime-silicon` (often requires Python 3.10 for older versions like 1.13.1):**
+    The original `README` noted that `onnxruntime-silicon==1.13.1` was specific to Python 3.10. If you intend to use this exact version for CoreML:
+    ```bash
+    # Ensure you are using Python 3.10 if required by your chosen onnxruntime-silicon version
+    # After running setup_mac.sh and activating .venv:
+    # source .venv/bin/activate
+    
+    pip uninstall onnxruntime onnxruntime-gpu # Uninstall any existing onnxruntime
+    pip install onnxruntime-silicon==1.13.1   # Or your desired version
+    
+    # Then use ./run_mac_coreml.sh
+    ```
+    Check the ONNX Runtime documentation for the latest recommended packages for Apple Silicon.
+*   **For MPS with ONNX Runtime:** This may require a specific build or version of `onnxruntime`. Consult the ONNX Runtime documentation. For PyTorch-based operations (like the Face Enhancer or Hair Segmenter if they were PyTorch native and not ONNX), PyTorch should automatically try to use MPS on compatible Apple Silicon hardware if available.
+*   **User Interface (Tkinter):** If you encounter errors related to `_tkinter` not being found when launching the UI, ensure your Python installation supports Tk. For Python installed via Homebrew, this is usually `python-tk` (e.g., `brew install python-tk@3.9` or `brew install python-tk@3.10`, matching your Python version).
 
 ** In case something goes wrong and you need to reinstall the virtual environment **
 

modules/globals.py

@@ -41,3 +41,4 @@ show_mouth_mask_box = False
 mask_feather_ratio = 8
 mask_down_size = 0.50
 mask_size = 1
+enable_hair_swapping = True  # Default state for enabling/disabling hair swapping

modules/hair_segmenter.py

@@ -4,6 +4,11 @@ from PIL import Image
 from transformers import SegformerImageProcessor, SegformerForSemanticSegmentation
 import cv2 # Imported for BGR to RGB conversion, though PIL can also do it.
 
+# Global variables for caching
+HAIR_SEGMENTER_PROCESSOR = None
+HAIR_SEGMENTER_MODEL = None
+MODEL_NAME = "isjackwild/segformer-b0-finetuned-segments-skin-hair-clothing"
+
 def segment_hair(image_np: np.ndarray) -> np.ndarray:
     """
     Segments hair from an image.
@@ -14,15 +19,41 @@ def segment_hair(image_np: np.ndarray) -> np.ndarray:
     Returns:
         NumPy array representing the binary hair mask.
     """
-    processor = SegformerImageProcessor.from_pretrained("isjackwild/segformer-b0-finetuned-segments-skin-hair-clothing")
-    model = SegformerForSemanticSegmentation.from_pretrained("isjackwild/segformer-b0-finetuned-segments-skin-hair-clothing")
+    global HAIR_SEGMENTER_PROCESSOR, HAIR_SEGMENTER_MODEL
+
+    if HAIR_SEGMENTER_PROCESSOR is None or HAIR_SEGMENTER_MODEL is None:
+        print(f"Loading hair segmentation model and processor ({MODEL_NAME}) for the first time...")
+        try:
+            HAIR_SEGMENTER_PROCESSOR = SegformerImageProcessor.from_pretrained(MODEL_NAME)
+            HAIR_SEGMENTER_MODEL = SegformerForSemanticSegmentation.from_pretrained(MODEL_NAME)
+            # Optional: Move model to GPU if available and if other models use GPU
+            # if torch.cuda.is_available():
+            #     HAIR_SEGMENTER_MODEL = HAIR_SEGMENTER_MODEL.to('cuda')
+            #     print("Hair segmentation model moved to GPU.")
+            print("Hair segmentation model and processor loaded successfully.")
+        except Exception as e:
+            print(f"Failed to load hair segmentation model/processor: {e}")
+            # Return an empty mask compatible with expected output shape (H, W)
+            return np.zeros((image_np.shape[0], image_np.shape[1]), dtype=np.uint8)
+
+    # Ensure processor and model are loaded before proceeding
+    if HAIR_SEGMENTER_PROCESSOR is None or HAIR_SEGMENTER_MODEL is None:
+        print("Error: Hair segmentation models are not available.")
+        return np.zeros((image_np.shape[0], image_np.shape[1]), dtype=np.uint8)
 
     # Convert BGR (OpenCV) to RGB (PIL)
     image_rgb = cv2.cvtColor(image_np, cv2.COLOR_BGR2RGB)
     image_pil = Image.fromarray(image_rgb)
 
-    inputs = processor(images=image_pil, return_tensors="pt")
-    outputs = model(**inputs)
+    inputs = HAIR_SEGMENTER_PROCESSOR(images=image_pil, return_tensors="pt")
+    
+    # Optional: Move inputs to GPU if model is on GPU
+    # if HAIR_SEGMENTER_MODEL.device.type == 'cuda':
+    #     inputs = inputs.to(HAIR_SEGMENTER_MODEL.device)
+
+    with torch.no_grad(): # Important for inference
+        outputs = HAIR_SEGMENTER_MODEL(**inputs)
+    
     logits = outputs.logits  # Shape: batch_size, num_labels, height, width
 
     # Upsample logits to original image size
@@ -33,12 +64,10 @@ def segment_hair(image_np: np.ndarray) -> np.ndarray:
         align_corners=False
     )
 
-    segmentation_map = upsampled_logits.argmax(dim=1).squeeze().cpu().numpy()
+    segmentation_map = upsampled_logits.argmax(dim=1).squeeze().cpu().numpy().astype(np.uint8)
 
     # Label 2 is for hair in this model
-    hair_mask = np.where(segmentation_map == 2, 255, 0).astype(np.uint8)
-
-    return hair_mask
+    return np.where(segmentation_map == 2, 255, 0).astype(np.uint8)
 
 if __name__ == '__main__':
     # This is a conceptual test.

modules/processors/frame/face_swapper.py

@@ -1,4 +1,4 @@
-from typing import Any, List
+from typing import Any, List, Optional, Tuple
 import cv2
 import insightface
 import threading
@@ -68,94 +68,133 @@ def get_face_swapper() -> Any:
     return FACE_SWAPPER
 
 
+def _prepare_warped_source_material_and_mask(
+    source_face_obj: Face, 
+    source_frame_full: Frame, 
+    matrix: np.ndarray, 
+    dsize: tuple # Built-in tuple is fine here for parameter type
+) -> Tuple[Optional[Frame], Optional[Frame]]:
+    """
+    Prepares warped source material (full image) and a combined (face+hair) mask for blending.
+    Returns (None, None) if essential masks cannot be generated.
+    """
+    # Generate Hair Mask
+    hair_only_mask_source_raw = segment_hair(source_frame_full)
+    if hair_only_mask_source_raw.ndim == 3 and hair_only_mask_source_raw.shape[2] == 3:
+        hair_only_mask_source_raw = cv2.cvtColor(hair_only_mask_source_raw, cv2.COLOR_BGR2GRAY)
+    _, hair_only_mask_source_binary = cv2.threshold(hair_only_mask_source_raw, 127, 255, cv2.THRESH_BINARY)
+
+    # Generate Face Mask
+    face_only_mask_source_raw = create_face_mask(source_face_obj, source_frame_full)
+    _, face_only_mask_source_binary = cv2.threshold(face_only_mask_source_raw, 127, 255, cv2.THRESH_BINARY)
+
+    # Combine Face and Hair Masks
+    if face_only_mask_source_binary.shape != hair_only_mask_source_binary.shape:
+        logging.warning("Resizing hair mask to match face mask for source during preparation.")
+        hair_only_mask_source_binary = cv2.resize(
+            hair_only_mask_source_binary, 
+            (face_only_mask_source_binary.shape[1], face_only_mask_source_binary.shape[0]), 
+            interpolation=cv2.INTER_NEAREST
+        )
+    
+    actual_combined_source_mask = cv2.bitwise_or(face_only_mask_source_binary, hair_only_mask_source_binary)
+    actual_combined_source_mask_blurred = cv2.GaussianBlur(actual_combined_source_mask, (5, 5), 3)
+
+    # Warp the Combined Mask and Full Source Material
+    warped_full_source_material = cv2.warpAffine(source_frame_full, matrix, dsize)
+    warped_combined_mask_temp = cv2.warpAffine(actual_combined_source_mask_blurred, matrix, dsize)
+    _, warped_combined_mask_binary_for_clone = cv2.threshold(warped_combined_mask_temp, 127, 255, cv2.THRESH_BINARY)
+    
+    return warped_full_source_material, warped_combined_mask_binary_for_clone
+
+
+def _blend_material_onto_frame(
+    base_frame: Frame, 
+    material_to_blend: Frame, 
+    mask_for_blending: Frame
+) -> Frame:
+    """
+    Blends material onto a base frame using a mask.
+    Uses seamlessClone if possible, otherwise falls back to simple masking.
+    """
+    x, y, w, h = cv2.boundingRect(mask_for_blending)
+    output_frame = base_frame # Start with base, will be modified by blending
+
+    if w > 0 and h > 0:
+        center = (x + w // 2, y + h // 2)
+        
+        if material_to_blend.shape == base_frame.shape and \
+           material_to_blend.dtype == base_frame.dtype and \
+           mask_for_blending.dtype == np.uint8:
+            try:
+                # Important: seamlessClone modifies the first argument (dst) if it's the same as the output var
+                # So, if base_frame is final_swapped_frame, it will be modified in place.
+                # If we want to keep base_frame pristine, it should be base_frame.copy() if it's also final_swapped_frame.
+                # Given final_swapped_frame is already a copy of swapped_frame at this point, this is fine.
+                output_frame = cv2.seamlessClone(material_to_blend, base_frame, mask_for_blending, center, cv2.NORMAL_CLONE)
+            except cv2.error as e:
+                logging.warning(f"cv2.seamlessClone failed: {e}. Falling back to simple blending.")
+                boolean_mask = mask_for_blending > 127 
+                output_frame[boolean_mask] = material_to_blend[boolean_mask]
+        else:
+            logging.warning("Mismatch in shape/type for seamlessClone. Falling back to simple blending.")
+            boolean_mask = mask_for_blending > 127
+            output_frame[boolean_mask] = material_to_blend[boolean_mask]
+    else:
+        logging.info("Warped mask for blending is empty. Skipping blending.")
+    
+    return output_frame
+
+
 def swap_face(source_face_obj: Face, target_face: Face, source_frame_full: Frame, temp_frame: Frame) -> Frame:
     face_swapper = get_face_swapper()
 
-    # Apply the face swap
-    swapped_frame = face_swapper.get(
-        temp_frame, target_face, source_face_obj, paste_back=True
-    )
+    # Apply the base face swap
+    swapped_frame = face_swapper.get(temp_frame, target_face, source_face_obj, paste_back=True)
+    final_swapped_frame = swapped_frame # Initialize with the base swap. Copy is made only if needed.
 
-    final_swapped_frame = swapped_frame.copy() # Initialize final_swapped_frame
+    if modules.globals.enable_hair_swapping:
+        if not (source_face_obj.kps is not None and \
+                target_face.kps is not None and \
+                source_face_obj.kps.shape[0] >= 3 and \
+                target_face.kps.shape[0] >= 3):
+            logging.warning(
+                f"Skipping hair blending due to insufficient keypoints. "
+                f"Source kps: {source_face_obj.kps.shape if source_face_obj.kps is not None else 'None'}, "
+                f"Target kps: {target_face.kps.shape if target_face.kps is not None else 'None'}."
+            )
+        else:
+            source_kps_float = source_face_obj.kps.astype(np.float32)
+            target_kps_float = target_face.kps.astype(np.float32)
+            matrix, _ = cv2.estimateAffinePartial2D(source_kps_float, target_kps_float, method=cv2.LMEDS)
 
-    # START of Hair Blending Logic
-    if source_face_obj.kps is not None and target_face.kps is not None and source_face_obj.kps.shape[0] >=2 and target_face.kps.shape[0] >=2 : # kps are 5x2 landmarks
-        hair_only_mask_source = segment_hair(source_frame_full)
-
-        # Ensure kps are float32 for estimateAffinePartial2D
-        source_kps_float = source_face_obj.kps.astype(np.float32)
-        target_kps_float = target_face.kps.astype(np.float32)
-
-        # b. Estimate Transformation Matrix
-        # Using LMEDS for robustness
-        matrix, _ = cv2.estimateAffinePartial2D(source_kps_float, target_kps_float, method=cv2.LMEDS)
-
-        if matrix is not None:
-            # c. Warp Source Hair and its Mask
-            dsize = (temp_frame.shape[1], temp_frame.shape[0]) # width, height
-
-            # Ensure hair_only_mask_source is 8-bit single channel
-            if hair_only_mask_source.ndim == 3 and hair_only_mask_source.shape[2] == 3:
-                hair_only_mask_source_gray = cv2.cvtColor(hair_only_mask_source, cv2.COLOR_BGR2GRAY)
+            if matrix is None:
+                logging.warning("Failed to estimate affine transformation matrix for hair. Skipping hair blending.")
             else:
-                hair_only_mask_source_gray = hair_only_mask_source
-            
-            # Threshold to ensure binary mask for warping
-            _, hair_only_mask_source_binary = cv2.threshold(hair_only_mask_source_gray, 127, 255, cv2.THRESH_BINARY)
-
-            warped_hair_mask = cv2.warpAffine(hair_only_mask_source_binary, matrix, dsize)
-            warped_source_hair_image = cv2.warpAffine(source_frame_full, matrix, dsize)
-
-            # d. Color Correct Warped Source Hair
-            # Using swapped_frame (face-swapped output) as the target for color correction
-            color_corrected_warped_hair = apply_color_transfer(warped_source_hair_image, swapped_frame)
-
-            # e. Blend Hair onto Swapped Frame
-            # Ensure warped_hair_mask is binary (0 or 255) after warping
-            _, warped_hair_mask_binary = cv2.threshold(warped_hair_mask, 127, 255, cv2.THRESH_BINARY)
-
-            # Preferred: cv2.seamlessClone
-            x, y, w, h = cv2.boundingRect(warped_hair_mask_binary)
-            if w > 0 and h > 0:
-                center = (x + w // 2, y + h // 2)
-                # seamlessClone expects target image, source image, mask, center, method
-                # The mask should be single channel 8-bit.
-                # The source (color_corrected_warped_hair) and target (swapped_frame) should be 8-bit 3-channel.
+                dsize = (temp_frame.shape[1], temp_frame.shape[0]) # width, height
                 
-                # Check if swapped_frame is suitable for seamlessClone (it should be the base)
-                # Ensure color_corrected_warped_hair is also 8UC3
-                if color_corrected_warped_hair.shape == swapped_frame.shape and \
-                   color_corrected_warped_hair.dtype == swapped_frame.dtype and \
-                   warped_hair_mask_binary.dtype == np.uint8:
-                    try:
-                        final_swapped_frame = cv2.seamlessClone(color_corrected_warped_hair, swapped_frame, warped_hair_mask_binary, center, cv2.NORMAL_CLONE)
-                    except cv2.error as e:
-                        logging.warning(f"cv2.seamlessClone failed: {e}. Falling back to simple blending.")
-                        # Fallback: Simple Blending (if seamlessClone fails)
-                        warped_hair_mask_3ch = cv2.cvtColor(warped_hair_mask_binary, cv2.COLOR_GRAY2BGR) > 0 # boolean mask
-                        final_swapped_frame[warped_hair_mask_3ch] = color_corrected_warped_hair[warped_hair_mask_3ch]
-                else:
-                    logging.warning("Mismatch in shape/type for seamlessClone. Falling back to simple blending.")
-                    # Fallback: Simple Blending
-                    warped_hair_mask_3ch = cv2.cvtColor(warped_hair_mask_binary, cv2.COLOR_GRAY2BGR) > 0
-                    final_swapped_frame[warped_hair_mask_3ch] = color_corrected_warped_hair[warped_hair_mask_3ch]
-            else:
-                # Mask is empty, no hair to blend, final_swapped_frame remains as is (copy of swapped_frame)
-                logging.info("Warped hair mask is empty. Skipping hair blending.")
-                # final_swapped_frame is already a copy of swapped_frame
-        else:
-            logging.warning("Failed to estimate affine transformation matrix for hair. Skipping hair blending.")
-            # final_swapped_frame is already a copy of swapped_frame
-    else:
-        if source_face_obj.kps is None or target_face.kps is None:
-            logging.warning("Source or target keypoints (kps) are None. Skipping hair blending.")
-        else:
-            logging.warning(f"Not enough keypoints for hair transformation. Source kps: {source_face_obj.kps.shape if source_face_obj.kps is not None else 'None'}, Target kps: {target_face.kps.shape if target_face.kps is not None else 'None'}. Skipping hair blending.")
-        # final_swapped_frame is already a copy of swapped_frame
-    # END of Hair Blending Logic
+                warped_material, warped_mask = _prepare_warped_source_material_and_mask(
+                    source_face_obj, source_frame_full, matrix, dsize
+                )
 
-    # f. Mouth Mask Logic
+                if warped_material is not None and warped_mask is not None:
+                    # Make a copy only now that we are sure we will modify it for hair.
+                    final_swapped_frame = swapped_frame.copy() 
+                    
+                    color_corrected_material = apply_color_transfer(warped_material, final_swapped_frame) # Use final_swapped_frame for color context
+                    
+                    final_swapped_frame = _blend_material_onto_frame(
+                        final_swapped_frame, 
+                        color_corrected_material, 
+                        warped_mask
+                    )
+    
+    # Mouth Mask Logic (operates on final_swapped_frame)
     if modules.globals.mouth_mask:
+        # If final_swapped_frame wasn't copied for hair, it needs to be copied now before mouth mask modification.
+        if final_swapped_frame is swapped_frame: # Check if it's still the same object
+            final_swapped_frame = swapped_frame.copy()
+
         # Create a mask for the target face
         face_mask = create_face_mask(target_face, temp_frame)
 
@@ -201,99 +240,91 @@ def process_frame(source_face_obj: Face, source_frame_full: Frame, temp_frame: F
 
 
 # process_frame_v2 needs to accept source_frame_full as well
+
+def _process_image_target_v2(source_frame_full: Frame, temp_frame: Frame) -> Frame:
+    if modules.globals.many_faces:
+        source_face_obj = default_source_face()
+        if source_face_obj:
+            for map_item in modules.globals.source_target_map:
+                target_face = map_item["target"]["face"]
+                temp_frame = swap_face(source_face_obj, target_face, source_frame_full, temp_frame)
+    else: # not many_faces
+        for map_item in modules.globals.source_target_map:
+            if "source" in map_item:
+                source_face_obj = map_item["source"]["face"]
+                target_face = map_item["target"]["face"]
+                temp_frame = swap_face(source_face_obj, target_face, source_frame_full, temp_frame)
+    return temp_frame
+
+def _process_video_target_v2(source_frame_full: Frame, temp_frame: Frame, temp_frame_path: str) -> Frame:
+    if modules.globals.many_faces:
+        source_face_obj = default_source_face()
+        if source_face_obj:
+            for map_item in modules.globals.source_target_map:
+                target_frames_data = [f for f in map_item.get("target_faces_in_frame", []) if f.get("location") == temp_frame_path]
+                for frame_data in target_frames_data:
+                    for target_face in frame_data.get("faces", []):
+                        temp_frame = swap_face(source_face_obj, target_face, source_frame_full, temp_frame)
+    else: # not many_faces
+        for map_item in modules.globals.source_target_map:
+            if "source" in map_item:
+                source_face_obj = map_item["source"]["face"]
+                target_frames_data = [f for f in map_item.get("target_faces_in_frame", []) if f.get("location") == temp_frame_path]
+                for frame_data in target_frames_data:
+                    for target_face in frame_data.get("faces", []):
+                        temp_frame = swap_face(source_face_obj, target_face, source_frame_full, temp_frame)
+    return temp_frame
+
+def _process_live_target_v2(source_frame_full: Frame, temp_frame: Frame) -> Frame:
+    detected_faces = get_many_faces(temp_frame)
+    if not detected_faces:
+        return temp_frame
+
+    if modules.globals.many_faces:
+        source_face_obj = default_source_face()
+        if source_face_obj:
+            for target_face in detected_faces:
+                temp_frame = swap_face(source_face_obj, target_face, source_frame_full, temp_frame)
+    else: # not many_faces (apply simple_map logic)
+        if not modules.globals.simple_map or \
+           not modules.globals.simple_map.get("target_embeddings") or \
+           not modules.globals.simple_map.get("source_faces"):
+            logging.warning("Simple map is not configured correctly. Skipping face swap.")
+            return temp_frame
+
+        target_embeddings = modules.globals.simple_map["target_embeddings"]
+        source_faces_from_map = modules.globals.simple_map["source_faces"]
+
+        if len(detected_faces) <= len(target_embeddings):
+            for detected_face in detected_faces:
+                closest_centroid_index, _ = find_closest_centroid(target_embeddings, detected_face.normed_embedding)
+                if closest_centroid_index < len(source_faces_from_map):
+                    source_face_obj_from_map = source_faces_from_map[closest_centroid_index]
+                    temp_frame = swap_face(source_face_obj_from_map, detected_face, source_frame_full, temp_frame)
+                else:
+                    logging.warning(f"Centroid index {closest_centroid_index} out of bounds for source_faces_from_map.")
+        else: # More detected faces than target embeddings in simple_map
+            detected_faces_embeddings = [face.normed_embedding for face in detected_faces]
+            for i, target_embedding in enumerate(target_embeddings):
+                if i < len(source_faces_from_map):
+                    closest_detected_face_index, _ = find_closest_centroid(detected_faces_embeddings, target_embedding)
+                    source_face_obj_from_map = source_faces_from_map[i]
+                    target_face_to_swap = detected_faces[closest_detected_face_index]
+                    temp_frame = swap_face(source_face_obj_from_map, target_face_to_swap, source_frame_full, temp_frame)
+                    # Optionally, remove the swapped detected face to prevent re-swapping if one source maps to multiple targets.
+                    # This depends on desired behavior. For now, simple independent mapping.
+                else:
+                    logging.warning(f"Index {i} out of bounds for source_faces_from_map in simple_map else case.")
+    return temp_frame
+
+
 def process_frame_v2(source_frame_full: Frame, temp_frame: Frame, temp_frame_path: str = "") -> Frame:
     if is_image(modules.globals.target_path):
-        if modules.globals.many_faces:
-            source_face_obj = default_source_face() # This function needs to be checked if it needs source_frame_full
-            if source_face_obj: # Ensure default_source_face actually returns a face
-                 for map_item in modules.globals.source_target_map: # Renamed map to map_item to avoid conflict
-                    target_face = map_item["target"]["face"]
-                    temp_frame = swap_face(source_face_obj, target_face, source_frame_full, temp_frame)
-
-        elif not modules.globals.many_faces:
-            for map_item in modules.globals.source_target_map: # Renamed map to map_item
-                if "source" in map_item:
-                    source_face_obj = map_item["source"]["face"]
-                    target_face = map_item["target"]["face"]
-                    temp_frame = swap_face(source_face_obj, target_face, source_frame_full, temp_frame)
-
+        return _process_image_target_v2(source_frame_full, temp_frame)
     elif is_video(modules.globals.target_path):
-        if modules.globals.many_faces:
-            source_face_obj = default_source_face() # This function needs to be checked
-            if source_face_obj:
-                for map_item in modules.globals.source_target_map: # Renamed map to map_item
-                    target_frames_data = [ # Renamed target_frame to target_frames_data
-                        f
-                        for f in map_item["target_faces_in_frame"]
-                        if f["location"] == temp_frame_path
-                    ]
-
-                    for frame_data in target_frames_data: # Renamed frame to frame_data
-                        for target_face in frame_data["faces"]:
-                            temp_frame = swap_face(source_face_obj, target_face, source_frame_full, temp_frame)
-
-        elif not modules.globals.many_faces:
-            for map_item in modules.globals.source_target_map: # Renamed map to map_item
-                if "source" in map_item:
-                    target_frames_data = [ # Renamed target_frame to target_frames_data
-                        f
-                        for f in map_item["target_faces_in_frame"]
-                        if f["location"] == temp_frame_path
-                    ]
-                    source_face_obj = map_item["source"]["face"]
-
-                    for frame_data in target_frames_data: # Renamed frame to frame_data
-                        for target_face in frame_data["faces"]:
-                            temp_frame = swap_face(source_face_obj, target_face, source_frame_full, temp_frame)
-
+        return _process_video_target_v2(source_frame_full, temp_frame, temp_frame_path)
     else: # This is the live cam / generic case
-        detected_faces = get_many_faces(temp_frame)
-        if modules.globals.many_faces:
-            if detected_faces:
-                source_face_obj = default_source_face() # This function needs to be checked
-                if source_face_obj:
-                    for target_face in detected_faces:
-                        temp_frame = swap_face(source_face_obj, target_face, source_frame_full, temp_frame)
-
-        elif not modules.globals.many_faces:
-            if detected_faces:
-                if len(detected_faces) <= len(
-                    modules.globals.simple_map["target_embeddings"]
-                ):
-                    for detected_face in detected_faces:
-                        closest_centroid_index, _ = find_closest_centroid(
-                            modules.globals.simple_map["target_embeddings"],
-                            detected_face.normed_embedding,
-                        )
-                        # Assuming simple_map["source_faces"] are Face objects
-                        # And default_source_face() logic might need to be more complex if source_frame_full is always from a single source_path
-                        source_face_obj_from_map = modules.globals.simple_map["source_faces"][closest_centroid_index]
-                        temp_frame = swap_face(
-                            source_face_obj_from_map, # This is source_face_obj
-                            detected_face,           # This is target_face
-                            source_frame_full,       # This is source_frame_full
-                            temp_frame,
-                        )
-                else:
-                    detected_faces_centroids = []
-                    for face in detected_faces:
-                        detected_faces_centroids.append(face.normed_embedding)
-                    i = 0
-                    for target_embedding in modules.globals.simple_map[
-                        "target_embeddings"
-                    ]:
-                        closest_centroid_index, _ = find_closest_centroid(
-                            detected_faces_centroids, target_embedding
-                        )
-                        source_face_obj_from_map = modules.globals.simple_map["source_faces"][i]
-                        temp_frame = swap_face(
-                            source_face_obj_from_map, # source_face_obj
-                            detected_faces[closest_centroid_index], # target_face
-                            source_frame_full, # source_frame_full
-                            temp_frame,
-                        )
-                        i += 1
-    return temp_frame
+        return _process_live_target_v2(source_frame_full, temp_frame)
 
 
 def process_frames(
@@ -353,30 +384,34 @@ def process_image(source_path: str, target_path: str, output_path: str) -> None:
         logging.error(f"Failed to read target image from {target_path}")
         return
 
+    # Read the original target frame once at the beginning
+    original_target_frame = cv2.imread(target_path)
+    if original_target_frame is None:
+        logging.error(f"Failed to read original target image from {target_path}")
+        return
+
+    result = None  # Initialize result
+
     if not modules.globals.map_faces:
         source_face_obj = get_one_face(source_img) # Use source_img here
         if not source_face_obj:
             logging.error(f"No face detected in source image {source_path}")
             return
-        result = process_frame(source_face_obj, source_img, target_frame)
-        cv2.imwrite(output_path, result)
-    else:
-        # map_faces == True for process_image
-        # process_frame_v2 expects source_frame_full as its first argument.
-        # The output_path is often the same as target_path initially for images.
-        # We read the target_frame (which will be modified)
-        target_frame_for_v2 = cv2.imread(output_path) # Or target_path, depending on desired workflow
-        if target_frame_for_v2 is None:
-            logging.error(f"Failed to read image for process_frame_v2 from {output_path}")
-            return
-
+        result = process_frame(source_face_obj, source_img, original_target_frame)
+    else: # map_faces is True
         if modules.globals.many_faces:
             update_status(
                 "Many faces enabled. Using first source image. Progressing...", NAME
             )
-        # Pass source_img (as source_frame_full) to process_frame_v2
-        result = process_frame_v2(source_img, target_frame_for_v2, target_path) # target_path as temp_frame_path hint
+        # process_frame_v2 takes the original target frame for processing.
+        # target_path is passed as temp_frame_path for consistency with process_frame_v2's signature,
+        # used for map lookups in video context but less critical for single images.
+        result = process_frame_v2(source_img, original_target_frame, target_path)
+
+    if result is not None:
         cv2.imwrite(output_path, result)
+    else:
+        logging.error(f"Processing image {target_path} failed, result was None.")
 
 
 def process_video(source_path: str, temp_frame_paths: List[str]) -> None:
@@ -745,113 +780,3 @@ def apply_color_transfer(source, target):
     source = (source - source_mean) * (target_std / source_std) + target_mean
 
     return cv2.cvtColor(np.clip(source, 0, 255).astype("uint8"), cv2.COLOR_LAB2BGR)
-
-
-def create_face_and_hair_mask(source_face: Face, source_frame: Frame) -> np.ndarray:
-    """
-    Creates a combined mask for the face and hair from the source image.
-    """
-    # 1. Generate the basic face mask (adapted from create_face_mask)
-    face_only_mask = np.zeros(source_frame.shape[:2], dtype=np.uint8)
-    landmarks = source_face.landmark_2d_106
-    if landmarks is not None:
-        landmarks = landmarks.astype(np.int32)
-
-        # Extract facial features (same logic as create_face_mask)
-        right_side_face = landmarks[0:16]
-        left_side_face = landmarks[17:32]
-        # right_eye = landmarks[33:42] # Not directly used for outline
-        right_eye_brow = landmarks[43:51]
-        # left_eye = landmarks[87:96] # Not directly used for outline
-        left_eye_brow = landmarks[97:105]
-
-        # Calculate forehead extension (same logic as create_face_mask)
-        right_eyebrow_top = np.min(right_eye_brow[:, 1])
-        left_eyebrow_top = np.min(left_eye_brow[:, 1])
-        eyebrow_top = min(right_eyebrow_top, left_eyebrow_top)
-
-        face_top = np.min([right_side_face[0, 1], left_side_face[-1, 1]])
-        # Ensure forehead_height is not negative if eyebrows are above the topmost landmark of face sides
-        forehead_height = max(0, face_top - eyebrow_top)
-        extended_forehead_height = int(forehead_height * 5.0)
-
-        forehead_left = right_side_face[0].copy()
-        forehead_right = left_side_face[-1].copy()
-        
-        # Ensure extended forehead points do not go into negative y values
-        forehead_left[1] = max(0, forehead_left[1] - extended_forehead_height)
-        forehead_right[1] = max(0, forehead_right[1] - extended_forehead_height)
-
-        face_outline = np.vstack(
-            [
-                [forehead_left],
-                right_side_face,
-                left_side_face[::-1],
-                [forehead_right],
-            ]
-        )
-
-        # Calculate padding (same logic as create_face_mask)
-        # Ensure face_outline has at least one point before calculating norm
-        if face_outline.shape[0] > 1:
-            padding = int(
-                np.linalg.norm(right_side_face[0] - left_side_face[-1]) * 0.05
-            )
-        else:
-            padding = 5 # Default padding if not enough points
-
-        hull = cv2.convexHull(face_outline)
-        hull_padded = []
-        center = np.mean(face_outline, axis=0).squeeze() # Squeeze to handle potential extra dim
-
-        # Ensure center is a 1D array for subtraction
-        if center.ndim > 1:
-            center = np.mean(center, axis=0)
-
-
-        for point_contour in hull:
-            point = point_contour[0] # cv2.convexHull returns points wrapped in an extra array
-            direction = point - center
-            norm_direction = np.linalg.norm(direction)
-            if norm_direction == 0: # Avoid division by zero if point is the center
-                unit_direction = np.array([0,0])
-            else:
-                unit_direction = direction / norm_direction
-            
-            padded_point = point + unit_direction * padding
-            hull_padded.append(padded_point)
-
-        if hull_padded: # Ensure hull_padded is not empty
-            hull_padded = np.array(hull_padded, dtype=np.int32)
-            cv2.fillConvexPoly(face_only_mask, hull_padded, 255)
-        else: # Fallback if hull_padded is empty (e.g. very few landmarks)
-            cv2.fillConvexPoly(face_only_mask, hull, 255) # Use unpadded hull
-
-
-        # Initial blur for face_only_mask is not strictly in the old one before combining,
-        # but can be applied here or after combining. Let's keep it like original for now.
-        # face_only_mask = cv2.GaussianBlur(face_only_mask, (5, 5), 3) # Original blur from create_face_mask
-
-    # 2. Generate the hair mask
-    # Ensure source_frame is contiguous, as some cv2 functions might require it.
-    source_frame_contiguous = np.ascontiguousarray(source_frame, dtype=np.uint8)
-    hair_mask_on_source = segment_hair(source_frame_contiguous)
-
-    # 3. Combine the masks
-    # Ensure masks are binary and of the same type for bitwise operations
-    _, face_only_mask_binary = cv2.threshold(face_only_mask, 127, 255, cv2.THRESH_BINARY)
-    _, hair_mask_on_source_binary = cv2.threshold(hair_mask_on_source, 127, 255, cv2.THRESH_BINARY)
-    
-    # Ensure shapes match. If not, hair_mask might be different. Resize if necessary.
-    # This should ideally not happen if segment_hair preserves dimensions.
-    if face_only_mask_binary.shape != hair_mask_on_source_binary.shape:
-        hair_mask_on_source_binary = cv2.resize(hair_mask_on_source_binary, 
-                                                (face_only_mask_binary.shape[1], face_only_mask_binary.shape[0]), 
-                                                interpolation=cv2.INTER_NEAREST)
-
-    combined_mask = cv2.bitwise_or(face_only_mask_binary, hair_mask_on_source_binary)
-
-    # 4. Apply Gaussian blur to the combined mask
-    combined_mask = cv2.GaussianBlur(combined_mask, (5, 5), 3)
-
-    return combined_mask

modules/ui.py

@@ -105,6 +105,7 @@ def save_switch_states():
         "show_fps": modules.globals.show_fps,
         "mouth_mask": modules.globals.mouth_mask,
         "show_mouth_mask_box": modules.globals.show_mouth_mask_box,
+        "enable_hair_swapping": modules.globals.enable_hair_swapping,
     }
     with open("switch_states.json", "w") as f:
         json.dump(switch_states, f)
@@ -129,6 +130,9 @@ def load_switch_states():
         modules.globals.show_mouth_mask_box = switch_states.get(
             "show_mouth_mask_box", False
         )
+        modules.globals.enable_hair_swapping = switch_states.get(
+            "enable_hair_swapping", True # Default to True if not found
+        )
     except FileNotFoundError:
         # If the file doesn't exist, use default values
         pass
@@ -284,6 +288,20 @@ def create_root(start: Callable[[], None], destroy: Callable[[], None]) -> ctk.C
     )
     show_fps_switch.place(relx=0.6, rely=0.75)
 
+    # Hair Swapping Switch (placed below "Show FPS" on the right column)
+    hair_swapping_value = ctk.BooleanVar(value=modules.globals.enable_hair_swapping)
+    hair_swapping_switch = ctk.CTkSwitch(
+        root,
+        text=_("Swap Hair"),
+        variable=hair_swapping_value,
+        cursor="hand2",
+        command=lambda: (
+            setattr(modules.globals, "enable_hair_swapping", hair_swapping_value.get()),
+            save_switch_states(),
+        )
+    )
+    hair_swapping_switch.place(relx=0.6, rely=0.80) # Adjusted rely from 0.75 to 0.80
+
     mouth_mask_var = ctk.BooleanVar(value=modules.globals.mouth_mask)
     mouth_mask_switch = ctk.CTkSwitch(
         root,
@@ -306,24 +324,26 @@ def create_root(start: Callable[[], None], destroy: Callable[[], None]) -> ctk.C
     )
     show_mouth_mask_box_switch.place(relx=0.6, rely=0.55)
 
+    # Adjusting placement of Start, Stop, Preview buttons due to new switch
     start_button = ctk.CTkButton(
         root, text=_("Start"), cursor="hand2", command=lambda: analyze_target(start, root)
     )
-    start_button.place(relx=0.15, rely=0.80, relwidth=0.2, relheight=0.05)
+    start_button.place(relx=0.15, rely=0.85, relwidth=0.2, relheight=0.05) # rely from 0.80 to 0.85
 
     stop_button = ctk.CTkButton(
         root, text=_("Destroy"), cursor="hand2", command=lambda: destroy()
     )
-    stop_button.place(relx=0.4, rely=0.80, relwidth=0.2, relheight=0.05)
+    stop_button.place(relx=0.4, rely=0.85, relwidth=0.2, relheight=0.05) # rely from 0.80 to 0.85
 
     preview_button = ctk.CTkButton(
         root, text=_("Preview"), cursor="hand2", command=lambda: toggle_preview()
     )
-    preview_button.place(relx=0.65, rely=0.80, relwidth=0.2, relheight=0.05)
+    preview_button.place(relx=0.65, rely=0.85, relwidth=0.2, relheight=0.05) # rely from 0.80 to 0.85
 
     # --- Camera Selection ---
+    # Adjusting placement of Camera selection due to new switch
     camera_label = ctk.CTkLabel(root, text=_("Select Camera:"))
-    camera_label.place(relx=0.1, rely=0.86, relwidth=0.2, relheight=0.05)
+    camera_label.place(relx=0.1, rely=0.91, relwidth=0.2, relheight=0.05) # rely from 0.86 to 0.91
 
     available_cameras = get_available_cameras()
     camera_indices, camera_names = available_cameras
@@ -342,7 +362,7 @@ def create_root(start: Callable[[], None], destroy: Callable[[], None]) -> ctk.C
             root, variable=camera_variable, values=camera_names
         )
 
-    camera_optionmenu.place(relx=0.35, rely=0.86, relwidth=0.25, relheight=0.05)
+    camera_optionmenu.place(relx=0.35, rely=0.91, relwidth=0.25, relheight=0.05) # rely from 0.86 to 0.91
 
     live_button = ctk.CTkButton(
         root,
@@ -362,16 +382,16 @@ def create_root(start: Callable[[], None], destroy: Callable[[], None]) -> ctk.C
             else "disabled"
         ),
     )
-    live_button.place(relx=0.65, rely=0.86, relwidth=0.2, relheight=0.05)
+    live_button.place(relx=0.65, rely=0.91, relwidth=0.2, relheight=0.05) # rely from 0.86 to 0.91
     # --- End Camera Selection ---
 
     status_label = ctk.CTkLabel(root, text=None, justify="center")
-    status_label.place(relx=0.1, rely=0.9, relwidth=0.8)
+    status_label.place(relx=0.1, rely=0.96, relwidth=0.8) # rely from 0.9 to 0.96
 
     donate_label = ctk.CTkLabel(
         root, text="Deep Live Cam", justify="center", cursor="hand2"
     )
-    donate_label.place(relx=0.1, rely=0.95, relwidth=0.8)
+    donate_label.place(relx=0.1, rely=0.99, relwidth=0.8) # rely from 0.95 to 0.99
     donate_label.configure(
         text_color=ctk.ThemeManager.theme.get("URL").get("text_color")
     )
@@ -880,7 +900,94 @@ def create_webcam_preview(camera_index: int):
     PREVIEW.deiconify()
 
     frame_processors = get_frame_processors_modules(modules.globals.frame_processors)
-    # source_image = None # Replaced by source_face_obj_for_cam
+    
+    # --- Source Image Loading and Validation (Moved before the loop) ---
+    source_face_obj_for_cam = None
+    source_frame_full_for_cam = None
+    source_frame_full_for_cam_map_faces = None
+
+    if not modules.globals.map_faces:
+        if not modules.globals.source_path:
+            update_status("Error: No source image selected for webcam mode.")
+            cap.release()
+            PREVIEW.withdraw()
+            while PREVIEW.state() != "withdrawn" and ROOT.winfo_exists():
+                ROOT.update_idletasks()
+                ROOT.update()
+                time.sleep(0.05)
+            return
+        if not os.path.exists(modules.globals.source_path):
+            update_status(f"Error: Source image not found at {modules.globals.source_path}")
+            cap.release()
+            PREVIEW.withdraw()
+            while PREVIEW.state() != "withdrawn" and ROOT.winfo_exists():
+                ROOT.update_idletasks()
+                ROOT.update()
+                time.sleep(0.05)
+            return
+        
+        source_frame_full_for_cam = cv2.imread(modules.globals.source_path)
+        if source_frame_full_for_cam is None:
+            update_status(f"Error: Could not read source image at {modules.globals.source_path}")
+            cap.release()
+            PREVIEW.withdraw()
+            while PREVIEW.state() != "withdrawn" and ROOT.winfo_exists():
+                ROOT.update_idletasks()
+                ROOT.update()
+                time.sleep(0.05)
+            return
+
+        source_face_obj_for_cam = get_one_face(source_frame_full_for_cam)
+        if source_face_obj_for_cam is None:
+            update_status(f"Error: No face detected in source image {modules.globals.source_path}")
+            # This error is less critical for stopping immediately, but we'll make it persistent too.
+            # The loop below will run, but processing for frames will effectively be skipped.
+            # For consistency in error handling, make it persistent.
+            cap.release()
+            PREVIEW.withdraw()
+            while PREVIEW.state() != "withdrawn" and ROOT.winfo_exists():
+                ROOT.update_idletasks()
+                ROOT.update()
+                time.sleep(0.05)
+            return
+    else: # modules.globals.map_faces is True
+        if not modules.globals.source_path:
+            update_status("Error: No global source image selected (for hair/background in map_faces mode).")
+            cap.release()
+            PREVIEW.withdraw()
+            while PREVIEW.state() != "withdrawn" and ROOT.winfo_exists():
+                ROOT.update_idletasks()
+                ROOT.update()
+                time.sleep(0.05)
+            return
+        if not os.path.exists(modules.globals.source_path):
+            update_status(f"Error: Source image (for hair/background) not found at {modules.globals.source_path}")
+            cap.release()
+            PREVIEW.withdraw()
+            while PREVIEW.state() != "withdrawn" and ROOT.winfo_exists():
+                ROOT.update_idletasks()
+                ROOT.update()
+                time.sleep(0.05)
+            return
+
+        source_frame_full_for_cam_map_faces = cv2.imread(modules.globals.source_path)
+        if source_frame_full_for_cam_map_faces is None:
+            update_status(f"Error: Could not read source image (for hair/background) at {modules.globals.source_path}")
+            cap.release()
+            PREVIEW.withdraw()
+            while PREVIEW.state() != "withdrawn" and ROOT.winfo_exists():
+                ROOT.update_idletasks()
+                ROOT.update()
+                time.sleep(0.05)
+            return
+        
+        if not modules.globals.source_target_map and not modules.globals.simple_map:
+            update_status("Warning: No face map defined for map_faces mode. Swapper may not work as expected.")
+            # This is a warning, not a fatal error for the preview window itself. Processing will continue.
+            # No persistent loop here, as it's a warning about functionality, not a critical load error.
+
+    # --- End Source Image Loading ---
+
     prev_time = time.time()
     fps_update_interval = 0.5
     frame_count = 0
@@ -907,80 +1014,28 @@ def create_webcam_preview(camera_index: int):
             )
 
         if not modules.globals.map_faces:
-            # Case 1: map_faces is False
-            source_face_obj_for_cam = None
-            source_frame_full_for_cam = None
-            if modules.globals.source_path and os.path.exists(modules.globals.source_path):
-                source_frame_full_for_cam = cv2.imread(modules.globals.source_path)
-                if source_frame_full_for_cam is not None:
-                    source_face_obj_for_cam = get_one_face(source_frame_full_for_cam)
-                    if source_face_obj_for_cam is None:
-                        update_status(f"Error: No face detected in source image at {modules.globals.source_path}")
-                        # Optional: could return here or allow running without a source face if some processors handle it
-                else:
-                    update_status(f"Error: Could not read source image at {modules.globals.source_path}")
-                    cap.release()
-                    PREVIEW.withdraw()
-                    return 
-            elif modules.globals.source_path: 
-                update_status(f"Error: Source image not found at {modules.globals.source_path}")
-                cap.release()
-                PREVIEW.withdraw()
-                return
-            else: 
-                update_status("Error: No source image selected for webcam mode.")
-                cap.release()
-                PREVIEW.withdraw()
-                return
-
-            for frame_processor in frame_processors:
-                if frame_processor.NAME == "DLC.FACE-ENHANCER":
-                    if modules.globals.fp_ui["face_enhancer"]:
-                        # Assuming face_enhancer's process_frame doesn't need source_face or source_frame_full
-                        temp_frame = frame_processor.process_frame(None, temp_frame) 
-                else:
-                    if source_face_obj_for_cam and source_frame_full_for_cam is not None:
+            # Case 1: map_faces is False - source_face_obj_for_cam and source_frame_full_for_cam are pre-loaded
+            if source_face_obj_for_cam and source_frame_full_for_cam is not None: # Check if valid after pre-loading
+                for frame_processor in frame_processors:
+                    if frame_processor.NAME == "DLC.FACE-ENHANCER":
+                        if modules.globals.fp_ui["face_enhancer"]:
+                            temp_frame = frame_processor.process_frame(None, temp_frame) 
+                    else:
                         temp_frame = frame_processor.process_frame(source_face_obj_for_cam, source_frame_full_for_cam, temp_frame)
-                    # else: temp_frame remains unchanged if source isn't ready
+            # If source image was invalid (e.g. no face), source_face_obj_for_cam might be None.
+            # In this case, the frame processors that need it will be skipped, effectively just showing the raw webcam frame.
+            # The error message is already persistent due to the pre-loop check.
         else:
-            # Case 2: map_faces is True
-            source_frame_full_for_cam_map_faces = None
-            if modules.globals.source_path and os.path.exists(modules.globals.source_path):
-                source_frame_full_for_cam_map_faces = cv2.imread(modules.globals.source_path)
-                if source_frame_full_for_cam_map_faces is None:
-                    update_status(f"Error: Could not read source image (for hair/background) at {modules.globals.source_path}")
-                    cap.release()
-                    PREVIEW.withdraw()
-                    return
-            elif modules.globals.source_path:
-                 update_status(f"Error: Source image (for hair/background) not found at {modules.globals.source_path}")
-                 cap.release()
-                 PREVIEW.withdraw()
-                 return
-            else:
-                update_status("Error: No global source image selected (for hair/background in map_faces mode).")
-                cap.release()
-                PREVIEW.withdraw()
-                return
-
-            # Also check if map is defined, though process_frame_v2 handles specific face mapping internally
-            if not modules.globals.source_target_map and not modules.globals.simple_map: # Check both map types
-                update_status("Error: No face map defined for map_faces mode.")
-                # This might not need a return if some processors can run without map
-                # but for face_swapper, it's likely needed.
-                # For now, we proceed and let process_frame_v2 handle it.
-
-            modules.globals.target_path = None # Standard for live mode
-            for frame_processor in frame_processors:
-                if frame_processor.NAME == "DLC.FACE-ENHANCER":
-                    if modules.globals.fp_ui["face_enhancer"]:
-                        # Pass source_frame_full_for_cam_map_faces for signature consistency
-                        # The enhancer can choose to ignore it if not needed.
+            # Case 2: map_faces is True - source_frame_full_for_cam_map_faces is pre-loaded
+            if source_frame_full_for_cam_map_faces is not None: # Check if valid after pre-loading
+                modules.globals.target_path = None # Standard for live mode
+                for frame_processor in frame_processors:
+                    if frame_processor.NAME == "DLC.FACE-ENHANCER":
+                        if modules.globals.fp_ui["face_enhancer"]:
+                            temp_frame = frame_processor.process_frame_v2(source_frame_full_for_cam_map_faces, temp_frame)
+                    else:
                         temp_frame = frame_processor.process_frame_v2(source_frame_full_for_cam_map_faces, temp_frame)
-                else:
-                    if source_frame_full_for_cam_map_faces is not None:
-                        temp_frame = frame_processor.process_frame_v2(source_frame_full_for_cam_map_faces, temp_frame)
-                    # else: temp_frame remains unchanged if global source for map_faces isn't ready
+            # If source_frame_full_for_cam_map_faces was invalid, error is persistent from pre-loop check.
 
         # Calculate and display FPS
         current_time = time.time()

modules/video_capture.py

@@ -50,7 +50,17 @@ class VideoCapturer:
                         continue
             else:
                 # Unix-like systems (Linux/Mac) capture method
-                self.cap = cv2.VideoCapture(self.device_index)
+                if platform.system() == "Darwin":  # macOS
+                    print("INFO: Attempting to use cv2.CAP_AVFOUNDATION for macOS camera.")
+                    self.cap = cv2.VideoCapture(self.device_index, cv2.CAP_AVFOUNDATION)
+                    if not self.cap or not self.cap.isOpened():
+                        print("WARN: cv2.CAP_AVFOUNDATION failed to open camera. Trying default backend for macOS.")
+                        # Release the failed attempt before trying again
+                        if self.cap:
+                            self.cap.release()
+                        self.cap = cv2.VideoCapture(self.device_index) # Fallback to default
+                else:  # Other Unix-like systems (e.g., Linux)
+                    self.cap = cv2.VideoCapture(self.device_index)
 
             if not self.cap or not self.cap.isOpened():
                 raise RuntimeError("Failed to open camera")

run-cuda.bat

@@ -1 +1,16 @@
-python run.py --execution-provider cuda
+@echo off
+set VENV_DIR=.venv
+
+:: Check if virtual environment exists
+if not exist "%VENV_DIR%\Scripts\activate.bat" (
+    echo Virtual environment '%VENV_DIR%' not found.
+    echo Please run setup_windows.bat first.
+    pause
+    exit /b 1
+)
+
+echo Activating virtual environment...
+call "%VENV_DIR%\Scripts\activate.bat"
+
+echo Starting the application with CUDA execution provider...
+python run.py --execution-provider cuda %*

run-directml.bat

@@ -1 +1,16 @@
-python run.py --execution-provider dml
+@echo off
+set VENV_DIR=.venv
+
+:: Check if virtual environment exists
+if not exist "%VENV_DIR%\Scripts\activate.bat" (
+    echo Virtual environment '%VENV_DIR%' not found.
+    echo Please run setup_windows.bat first.
+    pause
+    exit /b 1
+)
+
+echo Activating virtual environment...
+call "%VENV_DIR%\Scripts\activate.bat"
+
+echo Starting the application with DirectML execution provider...
+python run.py --execution-provider dml %*

run_mac.sh

@@ -0,0 +1,20 @@
+#!/usr/bin/env bash
+
+VENV_DIR=".venv"
+
+# Check if virtual environment exists
+if [ ! -d "$VENV_DIR" ]; then
+    echo "Virtual environment '$VENV_DIR' not found."
+    echo "Please run ./setup_mac.sh first to create the environment and install dependencies."
+    exit 1
+fi
+
+echo "Activating virtual environment..."
+source "$VENV_DIR/bin/activate"
+
+echo "Starting the application with CPU execution provider..."
+# Passes all arguments passed to this script (e.g., --source, --target) to run.py
+python3 run.py --execution-provider cpu "$@"
+
+# Deactivate after script finishes (optional, as shell context closes)
+# deactivate

run_mac_coreml.sh

@@ -0,0 +1,13 @@
+#!/usr/bin/env bash
+
+VENV_DIR=".venv"
+
+if [ ! -d "$VENV_DIR" ]; then
+    echo "Virtual environment '$VENV_DIR' not found."
+    echo "Please run ./setup_mac.sh first."
+    exit 1
+fi
+
+source "$VENV_DIR/bin/activate"
+echo "Starting the application with CoreML execution provider..."
+python3 run.py --execution-provider coreml "$@"

run_mac_cpu.sh

@@ -0,0 +1,13 @@
+#!/usr/bin/env bash
+
+VENV_DIR=".venv"
+
+if [ ! -d "$VENV_DIR" ]; then
+    echo "Virtual environment '$VENV_DIR' not found."
+    echo "Please run ./setup_mac.sh first."
+    exit 1
+fi
+
+source "$VENV_DIR/bin/activate"
+echo "Starting the application with CPU execution provider..."
+python3 run.py --execution-provider cpu "$@"

run_mac_mps.sh

@@ -0,0 +1,13 @@
+#!/usr/bin/env bash
+
+VENV_DIR=".venv"
+
+if [ ! -d "$VENV_DIR" ]; then
+    echo "Virtual environment '$VENV_DIR' not found."
+    echo "Please run ./setup_mac.sh first."
+    exit 1
+fi
+
+source "$VENV_DIR/bin/activate"
+echo "Starting the application with MPS execution provider (for Apple Silicon)..."
+python3 run.py --execution-provider mps "$@"

run_windows.bat

@@ -0,0 +1,20 @@
+@echo off
+set VENV_DIR=.venv
+
+:: Check if virtual environment exists
+if not exist "%VENV_DIR%\Scripts\activate.bat" (
+    echo Virtual environment '%VENV_DIR%' not found.
+    echo Please run setup_windows.bat first to create the environment and install dependencies.
+    pause
+    exit /b 1
+)
+
+echo Activating virtual environment...
+call "%VENV_DIR%\Scripts\activate.bat"
+
+echo Starting the application with CPU execution provider...
+:: Passes all arguments passed to this script to run.py
+python run.py --execution-provider cpu %*
+
+:: Optional: Deactivate after script finishes
+:: call deactivate

setup_mac.sh

@@ -0,0 +1,81 @@
+#!/usr/bin/env bash
+
+# Exit immediately if a command exits with a non-zero status.
+set -e
+
+echo "Starting macOS setup..."
+
+# 1. Check for Python 3
+echo "Checking for Python 3..."
+if ! command -v python3 &> /dev/null
+then
+    echo "Python 3 could not be found. Please install Python 3."
+    echo "You can often install it using Homebrew: brew install python"
+    exit 1
+fi
+
+# 2. Check Python version (>= 3.9)
+echo "Checking Python 3 version..."
+python3 -c 'import sys; exit(0) if sys.version_info >= (3,9) else exit(1)'
+if [ $? -ne 0 ]; then
+    echo "Python 3.9 or higher is required."
+    echo "Your version is: $(python3 --version)"
+    echo "Please upgrade your Python version. Consider using pyenv or Homebrew to manage Python versions."
+    exit 1
+fi
+echo "Python 3.9+ found: $(python3 --version)"
+
+# 3. Check for ffmpeg
+echo "Checking for ffmpeg..."
+if ! command -v ffmpeg &> /dev/null
+then
+    echo "WARNING: ffmpeg could not be found. This program requires ffmpeg for video processing."
+    echo "You can install it using Homebrew: brew install ffmpeg"
+    echo "Continuing with setup, but video processing might fail later."
+else
+    echo "ffmpeg found: $(ffmpeg -version | head -n 1)"
+fi
+
+# 4. Define virtual environment directory
+VENV_DIR=".venv"
+
+# 5. Create virtual environment
+if [ -d "$VENV_DIR" ]; then
+    echo "Virtual environment '$VENV_DIR' already exists. Skipping creation."
+else
+    echo "Creating virtual environment in '$VENV_DIR'..."
+    python3 -m venv "$VENV_DIR"
+fi
+
+# 6. Activate virtual environment (for this script's session)
+echo "Activating virtual environment..."
+source "$VENV_DIR/bin/activate"
+
+# 7. Upgrade pip
+echo "Upgrading pip..."
+pip install --upgrade pip
+
+# 8. Install requirements
+echo "Installing requirements from requirements.txt..."
+if [ -f "requirements.txt" ]; then
+    pip install -r requirements.txt
+else
+    echo "ERROR: requirements.txt not found. Cannot install dependencies."
+    # Deactivate on error if desired, or leave active for user to debug
+    # deactivate
+    exit 1
+fi
+
+echo ""
+echo "Setup complete!"
+echo ""
+echo "To activate the virtual environment in your terminal, run:"
+echo "  source $VENV_DIR/bin/activate"
+echo ""
+echo "After activating, you can run the application using:"
+echo "  python3 run.py [arguments]"
+echo "Or use one of the run_mac_*.sh scripts (e.g., ./run_mac_cpu.sh)."
+echo ""
+
+# Deactivate at the end of the script's execution (optional, as script session ends)
+# deactivate

setup_windows.bat

@@ -0,0 +1,79 @@
+@echo off
+echo Starting Windows setup...
+
+:: 1. Check for Python
+echo Checking for Python...
+python --version >nul 2>&1
+if errorlevel 1 (
+    echo Python could not be found in your PATH.
+    echo Please install Python 3 (3.10 or higher recommended) and ensure it's added to your PATH.
+    echo You can download Python from https://www.python.org/downloads/
+    pause
+    exit /b 1
+)
+
+:: Optional: Check Python version (e.g., >= 3.9 or >=3.10).
+:: This is a bit more complex in pure batch. For now, rely on user having a modern Python 3.
+:: The README will recommend 3.10.
+echo Found Python:
+python --version
+
+:: 2. Check for ffmpeg (informational)
+echo Checking for ffmpeg...
+ffmpeg -version >nul 2>&1
+if errorlevel 1 (
+    echo WARNING: ffmpeg could not be found in your PATH. This program requires ffmpeg for video processing.
+    echo Please download ffmpeg from https://ffmpeg.org/download.html and add it to your system's PATH.
+    echo (The README.md contains a link for a potentially easier ffmpeg install method using a PowerShell command)
+    echo Continuing with setup, but video processing might fail later.
+    pause
+) else (
+    echo ffmpeg found.
+)
+
+:: 3. Define virtual environment directory
+set VENV_DIR=.venv
+
+:: 4. Create virtual environment
+if exist "%VENV_DIR%\Scripts\activate.bat" (
+    echo Virtual environment '%VENV_DIR%' already exists. Skipping creation.
+) else (
+    echo Creating virtual environment in '%VENV_DIR%'...
+    python -m venv "%VENV_DIR%"
+    if errorlevel 1 (
+        echo Failed to create virtual environment. Please check your Python installation.
+        pause
+        exit /b 1
+    )
+)
+
+:: 5. Activate virtual environment (for this script's session)
+echo Activating virtual environment...
+call "%VENV_DIR%\Scripts\activate.bat"
+
+:: 6. Upgrade pip
+echo Upgrading pip...
+python -m pip install --upgrade pip
+
+:: 7. Install requirements
+echo Installing requirements from requirements.txt...
+if exist "requirements.txt" (
+    python -m pip install -r requirements.txt
+) else (
+    echo ERROR: requirements.txt not found. Cannot install dependencies.
+    pause
+    exit /b 1
+)
+
+echo.
+echo Setup complete!
+echo.
+echo To activate the virtual environment in your command prompt, run:
+echo   %VENV_DIR%\Scripts\activate.bat
+echo.
+echo After activating, you can run the application using:
+echo   python run.py [arguments]
+echo Or use one of the run-*.bat scripts (e.g., run-cuda.bat, run_windows.bat).
+echo.
+pause
+exit /b 0