Feature Extractors

Feature extractors are the core components that process your content and extract meaningful information. Mixpeek provides a variety of pre-built extractors and allows you to create custom ones.

​

Available Extractors

​

read

Processes and reads the content of files, enabling basic content extraction and analysis.

​

embed

Generates vector embeddings from content for similarity search and semantic analysis. Supports multiple embedding types:

• url: Creates embeddings from URL content
• text: Generates embeddings from text input
• keyword: Creates keyword-based embeddings

​

transcribe

Converts audio/video content into text, providing searchable transcriptions of spoken content.

​

describe

Generates natural language descriptions of content, particularly useful for images and videos.

​

detect

Performs object and entity detection in visual content:

• faces: Identifies and analyzes faces with configurable confidence thresholds
• Objects: Detects and labels visible objects
• Scenes: Identifies scene contexts

​

json_output

Structures extracted information into customizable JSON formats. Configure the output shape for:

• objects: Array of detected object strings
• scenes: Array of scene description strings
• Custom fields based on your needs