T

Rick StaaandRUFFY-369 845892959a feat(node): migrate Feature Bank node to new streampack (#3 )

This commit migrates the Feature Bank node, originally implemented by @RUFFY-369
on [RUFFY-369/ComfyUI-FeatureBank](https://github.com/RUFFY-369/ComfyUI-FeatureBank),
to the new streampack.

The Feature Bank node is a ComfyUI-native implementation of the method proposed in
[Looking Backward: Streaming Video-to-Video Translation with Feature Banks](https://arxiv.org/abs/2405.15757)
by Liang et al. It enhances the stream diffusion model by providing information from
previous frames, thereby improving temporal consistency and reducing flickering.

---------

Co-authored-by: RUFFY-369 <prakarshkaushik369@gmail.com>

2025-03-17 23:01:32 +01:00

src/feature_bank

feat(node): migrate Feature Bank node to new streampack (#3 )

2025-03-17 23:01:32 +01:00

README.md

docs: add node pack node categories explanation (#1 )

2025-03-13 17:55:28 +01:00

README.md

ComfyUI Node Pack

A suite of custom ComfyUI nodes for building real-time video and audio workflows using ComfyStream. This pack introduces nodes that enable live streaming, AI-enhanced processing, and real-time interaction with ComfyUI workflows.

Node Categories

The node pack is organized into three categories based on their function:

Foundation Nodes

Foundation Nodes introduce entirely new capabilities to ComfyUI, unlocking workflows that were previously impossible. These nodes typically:

Introduce new AI models or stable diffusion checkpoints.
Enable core functionality that can be further expanded with Light Nodes.

Note

No Foundation Nodes have been added yet.

Light Nodes

Light Nodes modify or enhance existing workflows by:

Tweaking parameters or intermediate values in Foundation Nodes.
Providing workflow insights or debug information.
Creating variations of existing functionalities for new use cases.

Note

No Light Nodes have been added yet.

Input/Output Nodes

These nodes extend ComfyUI's input and output capabilities, making real-time streaming workflows possible. They allow users to:

Load and process video, audio, or tensor data from external sources.
Stream outputs to Livepeer, WebRTC, or other real-time platforms.
Connect external data sources for AI-driven workflows.

Examples:

LoadTensor - Loads tensor data for real-time processing.
LoadAudioTensor - Processes real-time audio input.

Note

No additional Input/Output Nodes have been added yet.