Easy Apply InstantID(instantIDApply)

This node applies InstantID identity preservation inside an EasyUse pipeline. It takes a face image, a selected InstantID model, an optional ControlNet, and InsightFace-based face analysis to inject identity information into the model and produce positive/negative conditioning suitable for sampling.

Inputs

Required

• pipe (PIPE_LINE): The pipeline state containing the model, conditioning, and other sampling data. The node updates this pipeline with the InstantID adapter applied.
• image (IMAGE): The face image used to extract the identity embedding for InstantID.
• instantid_file (COMBO): Selects the InstantID model file to use from the available files.
• insightface (COMBO): The device backend used for InsightFace face analysis. Options: CPU, CUDA, ROCM. Choose the backend matching your available hardware.
• control_net_name (COMBO): Selects the ControlNet model file to use for structure/pose guidance alongside InstantID.
• cn_strength (FLOAT): Strength of the applied ControlNet. Default is 1.0, range 0.0 to 10.0.
• cn_soft_weights (FLOAT): Soft weights for the ControlNet contribution. Default is 1.0, range 0.0 to 1.0.
• weight (FLOAT): Overall strength of the InstantID identity adapter. Default is 0.8, range 0.0 to 5.0.
• start_at (FLOAT): The fraction of the denoising process at which the InstantID/ControlNet conditioning starts taking effect. Default is 0.0, range 0.0 to 1.0.
• end_at (FLOAT): The fraction of the denoising process at which the InstantID/ControlNet conditioning stops taking effect. Default is 1.0, range 0.0 to 1.0.
• noise (FLOAT): Amount of noise applied to the conditioning/identity embedding. Default is 0.35, range 0.0 to 1.0, step 0.05.

Optional

• image_kps (IMAGE): Optional keypoint/pose image used as ControlNet keypoint input for structure guidance.
• mask (MASK): Optional mask that limits the InstantID effect to a specific region of the image.
• control_net (CONTROL_NET): Optional preloaded ControlNet model. When provided, this can override the model selected by control_net_name.

Outputs

• PIPE_LINE: The updated pipeline state with InstantID conditioning applied.
• MODEL: The patched model with the InstantID adapter and ControlNet applied.
• CONDITIONING: Positive conditioning containing the InstantID identity information.
• CONDITIONING: Negative conditioning for use during sampling.

Usage Notes

• Use the positive and negative conditioning outputs when sampling with the returned model.
• The image should be a clear face image for reliable identity extraction.
• If image_kps is supplied, it can be used to transfer pose/structure information separately from the identity image.
• The start_at and end_at values let you control when during denoising the InstantID/ControlNet guidance is active, which is useful for balancing identity fidelity against overall composition.
• Selecting the correct insightface backend matters for performance and compatibility. Use CUDA or ROCM when GPU acceleration is available, or CPU as a fallback.