comfyui-node-datatypes

To map ComfyUI data types, define node schemas using IO types such as io.Image and io.Latent. This creates required mappings between python types and io.Data types to ensure consistent tensor handling and node behavior.

ComfyUI tensor formats support image, latent, conditioning, mask, audio, video, and 3D workflows. Reference tables document these data type definitions to ensure correct execution logic for both V3 and V1 APIs.

To convert latents to images, define your ComfyUI node schema using io.Latent for input and io.Image for output. Implement the execute method to consume and produce these defined tensor types.

Yes, the ComfyUI V3 API supports conditioning and model type definitions. It provides reference tables for MODEL, CLIP, VAE, and CONDITIONING types to handle tensor interactions across both V3 and V1 APIs.

Passing MASK tensors fails when node schemas lack correct io.Data type mappings. You must properly declare MASK inputs and outputs in the schema to ensure consistent tensor handling during model interactions.