Skip to main content
工作流应用

Video Parsing Node

The Video Parsing Node is designed for multimodal temporal analysis. It simultaneously performs speech transcription and keyframe extraction, generates content descriptions for each keyframe, and structurally aligns all audiovisual information along the timeline.

Usage Example

For example, to parse a video and generate a content summary: This workflow receives a video file URL as input. The Video Parsing Node extracts speech-to-text transcripts, keyframe visual descriptions, and timing information; then, a Large Language Model (LLM) Node generates a concise video content summary.

Node Integration

Method 1: Add via the + Button

  • Hover your mouse over any node on the canvas (e.g., the Start Node); a + button automatically appears on its right side.
  • Click the + button and select the desired node from the pop-up node list. The node will be automatically added to the canvas and connected to the upstream node.

Method 2: Drag-and-Drop from the Node Library

  • In the left-side node library, locate and select Data → Video Parsing.
  • Drag the node onto the canvas at your desired location, then drag a connection line from the edge of an upstream node to the Video Parsing Node.

Core Configuration

Input

Specify the video to be parsed. Supports either direct URL input or referencing an upstream node variable. Use the dropdown to select the input method:
MethodDescriptionUse Case
ReferenceReference an upstream node variableVideo URL (String type) from an upstream node, or a File-type video uploaded via the Start Node.
InputDirectly enter the video URLFixed or externally sourced video URLs.
Limitations:
  • Maximum file size: 512 MB per file.
  • Only single-file parsing is supported — batch or list-based parsing is not allowed.
  • Supported formats: .mp4, .mkv, .avi, .mov, .wmv.

Parser Selection

Parser TypeBillingDescription
LLM Audio-Visual ParserNot billed yetBuilt-in LLM-powered parsing; prompt customization is not supported.

Node Output

The Video Parsing Node outputs structured parsing results based on video content, including file type and time-aligned audio transcripts and video keyframe descriptions.
Variable NameTypeDescription
fileTypeStringFile extension (e.g., "mp4", "mkv").
segmentsArray<Object>Array of parsed segments, organized chronologically. Each element represents one scene segment.
Structure of each segments array element:
Field NameTypeDescription
audioFramesArray<Object>Audio frame array containing ASR details within this segment.
videoFramesArray<Object>Video frame array containing keyframe visual descriptions within this segment.
startTimeNumberSegment start time (in milliseconds).
endTimeNumberSegment end time (in milliseconds).
indexNumberSegment index, starting from 0.
Structure of each audioFrames array element:
Field NameTypeDescription
ASRInfoStringSpeech-to-text transcript (empty string if no audio).
startTimeNumberAudio frame start time (in milliseconds).
endTimeNumberAudio frame end time (in milliseconds).
Structure of each videoFrames array element:
Field NameTypeDescription
textInfoStringKeyframe visual description (scene text generated by the LLM).
fileUrlStringKeyframe image URL (signed screenshot URL with expiration).
startTimeNumberKeyframe start time (in milliseconds).
endTimeNumberKeyframe end time (in milliseconds).

Common Usage Patterns

Paired NodeTypical Use Case
Start Node → Video Parsing NodeStart Node provides either a video URL (String type) or uploaded video file (File type); Video Parsing Node parses the video content.
Video Parsing Node → LLM NodeVideo Parsing Node outputs structured data (e.g., ASR text and visual descriptions from segments); LLM Node generates summaries, answers questions, or extracts insights based on these results.
Video Parsing Node → Variable Processing NodeExtract specific fields from parsing results for format conversion or data processing (e.g., concatenate all textInfo values into a unified visual description).