ComfyUI PSD Smart Object Nodes

Custom ComfyUI nodes that understand Photoshop mockup files. Drop a PSD/PSB into ComfyUI/input, feed any image (from the usual Load Image node or by providing a path on disk), and the node will automatically project that image onto every smart-object layer you select. Perspective, rotation, and scale follow the transform data stored inside the PSD smart object, so the result matches what you would get when replacing the layer inside Photoshop.

Features

• PSD Smart Object Inspector – Lists each smart-object layer (name, id, bounding box, transform box). Use it to confirm the layer names you want to target.
• PSD Mockup Embedder – Uploads PSD/PSB files directly from the node, reuses the PSD’s original clipping stack (shadows, highlights, rounded corners, etc.), projects your image into the smart objects, resizes the result to any width (default 800 px), and emits the composited tensor plus an auto-saved PNG.
• Works with PSD/PSB files, multiple smart objects, and either ComfyUI IMAGE tensors or file paths.

Installation

1. Copy this folder into ComfyUI/custom_nodes/comfyui_psd_smart_object.
2. (Optional) install the lightweight dependencies inside your ComfyUI environment:

   pip install -r requirements.txt
   

3. Restart ComfyUI so it can discover the new nodes.

Usage

1. Place PSD/PSB mockups under ComfyUI/input/psd (or pass an absolute path).
2. On PSD Mockup Embedder, click “choose PSD to upload” (or drag a PSD onto the node). The bundled frontend script uploads the file into ComfyUI/input and fills the psd_file field automatically.
3. Load your artwork with any stock Load Image (or Upload Image) node and connect its IMAGE output to the embedder. Leave smart_object_names empty to target the first smart object automatically, or pass a comma-separated list to specify others.
4. Adjust output_width if you need a specific export size (set to 0 to keep the PSD’s native resolution), then trigger the workflow. The node outputs:
   • mockup_image: an IMAGE tensor that you can preview or post-process.
   • debug_info: JSON describing the layers that were replaced.
   • png_path: the auto-saved PNG under ComfyUI/output, so you no longer need an extra Save Image node to receive a PNG file.

Notes & Limitations

• The result is flattened, so advanced Photoshop blend modes (e.g., highlight or shadow layers that sit on top of the smart object) are baked into the background. Complex lighting stacks may still need Photoshop for final touch-ups.
• Smart objects without transform metadata fall back to their axis-aligned bounding boxes.
• Perspective warping depends on OpenCV. If ComfyUI logs an import error, install opencv-python-headless in the same environment.

Requirements

See requirements.txt. ComfyUI already brings Torch and Pillow, so the additional runtime dependencies are:

• psd-tools – parses PSD/PSB files and exposes smart-object metadata
• opencv-python-headless – handles the perspective warp
• numpy – tensor and array glue code

Development

The nodes live in psd_mockup_node.py. Run PYTHONPYCACHEPREFIX=/tmp/pycache python -m py_compile psd_mockup_node.py before shipping changes to make sure there are no syntax issues.