Get Started - Pixeltable Documentation

Overview

Build and share multimodal AI datasets without managing infrastructure. Work with your images, videos, audio, and documents through a unified Python API - process them with AI models, create embeddings, and publish your results for team collaboration or public research.

Quick Start

Requirements: Pixeltable >= 0.4.24 Replicate a dataset:

import pixeltable as pxt

coco_copy = pxt.replicate(
    remote_uri='pxt://pixeltable:fiftyone/coco_mini_2017',
    local_path='coco-copy'
)

Replicas are read-only locally, but you can query them, perform similarity searches, update them with pull(), or create independent copies. Publish your datasets (requires account and API key from signin.pixeltable.com):

pxt.publish(
    source='my-table',
    destination_uri='pxt://username/my-dataset'
)

After publishing, use push() to update the cloud replica with local changes. Access defaults to private; add access='public' to make it publicly accessible. Learn more in the Data Sharing Guide.