Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.comfy.org/llms.txt

Use this file to discover all available pages before exploring further.

This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHub
The Sampler AR Video node provides a specialized sampling method for autoregressive video models, such as those using Causal Forcing or Self-Forcing techniques. It manages all parameters related to the autoregressive (AR) loop directly within the workflow, making it easy to configure how the model generates video frames one step at a time.

Inputs

ParameterData TypeRequiredRangeDescription
num_frame_per_blockINTYes1 to 64Frames per autoregressive block. A value of 1 means the model generates one frame at a time (framewise), while a value of 3 means it generates three frames together (chunkwise). This setting must match the checkpoint’s training mode. Default: 1.

Outputs

Output NameData TypeDescription
SAMPLERSAMPLERA configured sampler object that uses the “ar_video” sampling function with the specified autoregressive parameters.

Source fingerprint (SHA-256): 7f722791f55a4d8bbda1c00becf25c42e67e3a487712fa674ac118e978cf9812