Wan Camera Embedding Advanced

An extended Wan camera embedding node with independent translation and rotation controls.

Installation

Clone this repository into ComfyUI/custom_nodes:

git clone https://github.com/thepororo/ComfyUI-WanCameraAdvanced.git

Restart ComfyUI after installation.

The optional Wan Camera Multi-Frame Reference node requires the wan22fmlf custom node. The camera embedding, path chaining, and visualizer nodes do not require it.

Nodes

The node appears at:

model/conditioning/wan/camera > Wan Camera Embedding Advanced

Connect camera_embedding to the camera_conditions input of WanCameraImageToVideo.

Wan Camera Embedding Visualizer converts native or advanced WAN_CAMERA_EMBEDDING data into an IMAGE sequence:

• ray_direction: maps the XYZ ray direction to RGB.
• ray_moment: maps the XYZ Plucker moment to RGB with sequence-wide scaling.
• frame_stride: outputs every Nth source frame to reduce preview memory.

Multiple Wan Camera Embedding Advanced nodes can be chained by connecting camera_path to the next node's optional previous_camera_path input. The next segment starts from the previous segment's final camera transform. The shared boundary frame is included only once, so chaining two 81-frame segments produces 161 frames.

Wan Camera Multi-Frame Reference combines a camera embedding with explicit start, middle, and end reference images. It reuses the installed wan22fmlf/Wan Multi-Frame Reference implementation and adds camera_conditions without replacing its latent image or mask conditioning. For an 81-frame video, use middle_frame = 40.

length must be 1 plus a multiple of 4, such as 81.

Suggested starting values

• Yaw right: rotate_y_deg = 15, all other motion values 0
• Truck right: translate_x = 0.3, all other motion values 0
• Tracking combination: translate_x = 0.2, rotate_y_deg = -8

Large rotations reveal parts of a scene that are not present in the source image. Use perspective-consistent middle and end reference images when attempting rotations above roughly 45 degrees.

License

GPL-3.0

Development

Designed and maintained by thepororo, with implementation assistance from OpenAI Codex. AI-assisted changes are reviewed and validated locally before publication.