ComfyUI Node
Landmark Facial Aligner
A ComfyUI node in ComfyUI-JH-PixelPro/geometry with 4 inputs and 2 outputs.
Landmark Facial Aligner
- image
- image_aligned
- inverse_matrix_json
◄landmarks[[820, 650], [1420, 640], [1120, 900], [930, 1150], [1310, 1145]]►
◄target_size1024►
◄padding0.20►
CategoryComfyUI-JH-PixelPro/geometry
Inputs (4)
| Name | Type | Default | Description |
|---|---|---|---|
| image | IMAGE | — | |
| landmarks | STRING | [[820, 650], [1420, 640], [1120, 900], [930, 1150], [1310, 1145]] | 5-point landmark JSON in order [L-eye, R-eye, nose, L-mouth, R-mouth]. Pixel-absolute or normalized (auto-detected: values ≤ 1.5 treated as normalized). Shape 5x2 for single image or Bx5x2 for batch. |
| target_size | INT | 1024512–1024 | Square output size in pixels (512 / 768 / 1024). Smaller is faster; 1024 is SDXL-friendly. |
| padding | FLOAT | 0.200–0.5 | Ratio of canonical frame reserved around the face (hair/chin room). 0 = tight crop, 0.5 = half-frame padding. |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| image_aligned | IMAGE | — |
| inverse_matrix_json | STRING | — |