Load CameraCtrl Poses (Path) ππ π β‘
Loading a camera trajectory from a file instead of building one by hand
- CAMERACTRL_POSES
Displayed as Load CameraCtrl Poses (Path), this is the simpler of the pack's two ways to get a CAMERACTRL_POSES object for CameraCtrl. Where ADE_CameraPoseCombo lets you build a camera move from dropdown segments inside ComfyUI, this node just reads an existing pose file straight off disk - useful if you already have (or found) a pose trajectory file rather than composing one by hand, matching the file-based pose format CameraCtrl's original research uses.
Know the prerequisites before this node does anything useful: per the README, CameraCtrl needs the pruned model CameraCtrl_pruned.safetensors, an AnimateDiff SD1.5 motion model trained for v3 specifically, and it's Gen2-only. This node's output only matters once those other pieces are in place - on its own, it's just a file read.
How it works
It parses a text file at file_path containing a camera-pose trajectory (the kind of frame-by-frame camera parameter sequence CameraCtrl's research release format uses) and returns it as a CAMERACTRL_POSES object, ready to feed ADE_ApplyAnimateDiffModelWithCameraCtrl. No editing or composition happens here - whatever trajectory is in the file is what you get.
The inputs and outputs that matter
There's exactly one input, and it's optional: file_path (STRING, default a placeholder X://path/to/pose_file.txt - you're expected to overwrite this with a real path). Output is CAMERACTRL_POSES, which only fits the cameractrl_poses input on ADE_ApplyAnimateDiffModelWithCameraCtrl - nowhere else in the pack consumes this type.
How to install it
Standard for the pack - ComfyUI Manager, search AnimateDiff Evolved by Kosinkadink, or:
cd ComfyUI/custom_nodes && git clone https://github.com/Kosinkadink/ComfyUI-AnimateDiff-Evolved
Beyond the pack itself, you need the pose file this node reads - either one you've authored to match CameraCtrl's expected format, or one shared by someone else's CameraCtrl workflow. You'll also need the CameraCtrl model (CameraCtrl_pruned.safetensors, linked from the README) and a v3 SD1.5 motion model to actually use the poses this node loads.
Common issues & troubleshooting
Left file_path on the placeholder value. This is the single most likely reason nothing happens - the default is a literal example path (X://path/to/pose_file.txt), not a real file. Point it at an actual pose file on your system.
File exists but the node errors or produces nonsense output. The file needs to match CameraCtrl's expected pose-trajectory format specifically - a text file that isn't in that format, or one built for a different camera-control system entirely, won't parse into something meaningful even if the file path itself is valid.
Not sure where to get a pose file in the first place, or whether to use this node vs. ADE_CameraPoseCombo. If you don't already have a specific pose file you're trying to load, ADE_CameraPoseCombo is the more approachable path - it builds a CAMERACTRL_POSES object from in-graph dropdowns (pan, tilt, zoom, roll segments) without needing an external file at all. Reach for this node specifically when you have a pose file already, from your own authoring or someone else's shared workflow.
CameraCtrl overall is a niche, setup-heavy corner of an already-niche pack - the community rarely discusses it, so treat troubleshooting here as mostly mechanical (get the model files and the file path right) rather than expecting much crowd-sourced advice if something goes wrong.
Inputs (1)
| Name | Type | Default | Description |
|---|---|---|---|
| file_pathopt | STRING | X://path/to/pose_file.txt | β |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| CAMERACTRL_POSES | CAMERACTRL_POSES | β |