controlnet-pose

Pose-conditioned image generation uses OpenPose, DWPose, depth, or canny control images to constrain model output. This Skill routes those references to the Z-Image Turbo ControlNet LoRA, ensuring the generated character matches your supplied pose skeleton or depth map.

Yes, depth and canny references can be used for pose-locked image generation. The Skill applies these control images alongside skeleton references to the appropriate RunComfy workflow, constraining the structural composition of the generated image or video.

Yes, the RunComfy CLI is required. Motion transfer and character animation routing requires authenticated CLI access, publicly reachable media URLs for your reference files, and valid model-specific input schemas to execute properly.

Skeleton-based pose conditioning works by extracting a pose map from a reference performance and applying it to a target character. The Skill routes this to Kling Motion Control, transferring the reference motion and blocking onto the generated character video.

Limitations include requiring publicly reachable media URLs and valid model-specific input schemas. Character animation routing depends on the RunComfy CLI and authenticated access, meaning local file paths without public hosting will not execute correctly.