Extract
Pulls information out of the text, image, audio or video.
The extract method allows you to pull data out of various modalities. Depending on the modality, different techniques can be applied:
- image: optical character recognition, object detection, etc.
- audio: transcription, diarization, etc.
- video: object detection, scene detection, etc.
- text: named entity recognition, tokenization, etc.
Currently only the text modality is available for public use. Contact us to use the beta for image, video and audio.
Request
Optional indication of which model to use
The input URL, text, or base64 encoding of the object to extract.
The type of data to be extracted (e.g., image, audio, video, text).
Note that modality is not required here. If you don’t include it, we’ll automatically detect how to process it. This is useful when it’s the first step in the pipeline and you don’t know what filetypes you want to process in advance.
Specify whether the input is a url, base64, or text.
Was this page helpful?