Indexes store and organize your video data, allowing you to group related videos. For example, you can create one index for all videos from a car race, then search for specific moments across all those videos in a single request.

Model configuration

When you create an index, configure which video understanding models process your videos and what modalities those models analyze.

Choose the models

Select one or both video understanding models based on your tasks:

To search videos: Enable Marengo.
To analyze videos and generate text: Enable Pegasus.

Choose the modalities

For each model, specify which modalities you want the platform to analyze:

Visual: Includes actions, objects, events, text through OCR, and brand logos.
Audio: Includes ambient sounds, music, and human speech.

Notes

The model configuration you specify applies to all videos you upload to the index.
You cannot change the model configuration after you create the index.

Create an index

Create a new index by calling the client.indexes.create function with the following parameters:

name: A descriptive name for your index
models: Your model configuration

Marengo

Pegasus

Marengo and Pegasus

This example creates an index with the Marengo model for video search, processing both visual and audio content:

1 index = client.indexes.create(
2     index_name="<YOUR_INDEX_NAME>",
3     models=[
4         IndexesCreateRequestModelsItem(
5             model_name="marengo2.7", model_options=["visual", "audio"]
6         )
7     ]
8 )
9 print(f"Created index: id={index.id}")

The response should look similar to the following one:

Created index: id=684018f7b4af41b9364d2b4d

Note that the response contains, among other information, a field named id, representing the unique identifier of your new index. Store this value to reference your index in future operations.