ComfyUI-VideoHelperSuite

This node supports multiple output formats, including common video formats (via FFmpeg) as well as GIF and WebP animation formats.

From a technical implementation perspective, this node internally maintains a state manager for inputs and outputs.

This node scans all supported audio format files in the ComfyUI input folder and generates a selectable file list.

This node provides extensive video processing options, allowing users to have fine-grained control over the loading process.

The main function of this node is to read video files from ComfyUI's input directory and decompress them frame by frame into image tensors.

This node supports loading various common image formats and features a built-in intelligent file sorting mechanism to ensure images are read in the correct order.

This node takes a tensor of type `IMAGE` as input, which typically represents a batch containing multiple image frames.

This node technically processes the input VHS_AUDIO data stream by invoking an FFmpeg subprocess.

The main function of this node is to parse video files into frame-by-frame images (or latent representations), while extracting audio tracks and detailed video information (such as frame rate, resolution, duration).

This node receives a tensor containing multiple images as input and uses two integer parameters to control the filtering logic.

This node technically supports loading multiple common audio formats, including WAV, MP3, OGG, M4A, and FLAC.

This node receives an input of type IMAGE, which is typically a tensor with a shape of `[batch_size, height, width, channels]`, representing a batch of images.

From a technical implementation perspective, this node receives a standard `LATENT` dictionary input, which typically contains key tensors such as `samples`.

This node receives pixel data of type IMAGE and a VAE model as inputs

This node technically processes the input path via the `load_video` function, so it can load not only common static image formats (such as PNG, JPG) but also handle sequences composed of multiple images.

The main function of this node is to serve as the starting point of the video processing workflow.

This node technically receives an AUDIO dictionary object containing sample rate and waveform data.

From a technical implementation perspective, this node receives an input of type `LATENT`, which is a dictionary containing the key `"samples"`.

In terms of technical implementation, this node accepts an input of type `VHS_VIDEOINFO`, a data structure typically generated by video loading nodes (such as Load Video) in the Video Helper Suite.

From a technical implementation perspective, this node receives a PyTorch tensor as input, typically with a shape of `[batch_size, height, width, channels]`.

This node determines the final output dimensions by comparing the sizes of two input latent tensors and following the user-selected merging strategy (such as match A, match B, match smaller, or match larger).

From a technical implementation perspective, this node takes a batch of images in PyTorch Tensor format as input and processes it through torch。

This node accepts a special data type `VHS_FILENAMES` as input, which is typically a structured list generated by other nodes in the same suite, such as the Load Video node.

From a technical implementation perspective, this node receives a MASK-type tensor as input, which typically represents a sequence composed of multiple frame masks.