EasyLoader (HunyuanDiT)(hunyuanDiTLoader)

Description

EasyLoader (HunyuanDiT) is a convenience loader for HunyuanDiT-based workflows. It loads the selected HunyuanDiT checkpoint and VAE, optionally applies a LoRA, and packages the model, latent settings, and prompt text into a PIPE_LINE for downstream sampling. It also exposes the loaded MODEL and VAE directly, making it useful both with EasyUse pipeline nodes and with standard ComfyUI sampler workflows.

Inputs

• ckpt_name (COMBO, required): The HunyuanDiT checkpoint to load.
• vae_name (COMBO, required): The VAE to load and use for decoding latents.
• lora_name (COMBO, required): The LoRA to apply to the loaded checkpoint.
• lora_model_strength (FLOAT, required, default=1, range -10 to 10): Strength of the LoRA applied to the model weights.
• lora_clip_strength (FLOAT, required, default=1, range -10 to 10): Strength of the LoRA applied to the text-encoder / CLIP side.
• resolution (COMBO, required, default "1024 x 1024"): Resolution preset for the generated latent.
• empty_latent_width (INT, required): Width of the empty latent to create for generation.
• empty_latent_height (INT, required): Height of the empty latent to create for generation.
• positive (STRING, required, default ""): Positive prompt text used to build conditioning.
• negative (STRING, required, default ""): Negative prompt text used to build conditioning. May be left empty.
• batch_size (INT, required, default=1, min=1, max=64): Number of latents to create in the batch.
• optional_lora_stack (LORA_STACK, optional): A stack of additional LoRAs to apply, with their own per-entry model and CLIP strengths.
• optional_controlnet_stack (CONTROL_NET_STACK, optional): A stack of ControlNets to apply while building the model pipeline.

Outputs

• PIPE_LINE: An EasyUse pipeline containing the loaded model, VAE, prompt conditioning, and latent settings for downstream EasyUse nodes.
• MODEL: The loaded HunyuanDiT model after LoRA and ControlNet modifications, ready for samplers.
• VAE: The loaded VAE, ready for decoding latents to images.

Usage Notes

• This node is intended as an all-in-one HunyuanDiT starting point. Connect PIPE_LINE to EasyUse pipeline samplers, or use MODEL and VAE directly in standard sampler workflows.
• The resolution combo is a convenience for choosing a common canvas size. Make sure empty_latent_width and empty_latent_height match the actual pixel dimensions you want to generate.
• LoRA can be supplied either as a single LoRA through lora_name with separate model and CLIP strengths, or as a stack through optional_lora_stack for combining multiple LoRAs in one pass.
• ControlNet settings are provided through optional_controlnet_stack; individual strength and preprocessor settings are carried by the stack entries rather than separate node inputs.
• positive and negative phrasing are carried inside the PIPE_LINE, so downstream sampler nodes can use them without requiring a separate conditioning loader.
• This node does not output a standalone CLIP/text-encoder model. The conditioning is built internally and stored in the PIPE_LINE output.