10-Minute Tour
Welcome to Pixeltable! In this tutorial, we'll survey how to create tables, populate them with data, and enhance them with built-in and user-defined transformations and AI operations.
Updated
103 notebooks and recipes from the Pixeltable docs, grouped the way the docs group them: open a notebook in Google Colab and run it.
Welcome to Pixeltable! In this tutorial, we'll survey how to create tables, populate them with data, and enhance them with built-in and user-defined transformations and AI operations.
Every multimodal AI app needs the same five things: store media, run models, index embeddings, serve endpoints, version everything.
Create a retrieval-augmented generation system that answers questions using your documents as context.
Build the six architectural patterns of agentic AI (prompt chaining, routing, parallelization, tool use, evaluator-optimizer, orchestrator-worker) and the ReAct and planning strategies as Pixeltable tables and computed columns.
Build visual similarity search to find images that look alike using OpenAI's CLIP model.
Detect objects in a video frame by frame: Pixeltable extracts the frames, runs YOLOX models on them and reassembles the frames into a video. Compare detection models and evaluate them against ground truth.
103 of 103 notebooks shown
Welcome to Pixeltable! In this tutorial, we'll survey how to create tables, populate them with data, and enhance them with built-in and user-defined transformations and AI operations.
Create and manage tables, insert, update and delete data, and filter and select rows with where(), select() and order_by().
This guide introduces one of Pixeltable's most essential and powerful concepts: computed columns.
Expressions are the basic building blocks of Pixeltable.
Pixeltable comes with a library of built-in functions and integrations, but sooner or later, you'll want to introduce some customized logic into your workflow.
Create an embedding index, query it with similarity search and keep it current as data changes. Covers several indexes on one column, custom embedding UDFs and precomputed embeddings.
In Pixeltable, all media data (videos, images, audio) resides in external files, and Pixeltable stores references to those.
Cookbooks
Every multimodal AI app needs the same five things: store media, run models, index embeddings, serve endpoints, version everything.
Set up API credentials for OpenAI, Anthropic, and other AI providers so Pixeltable can access them.
You need to iterate on transformation logic before running it on your entire dataset—especially for expensive operations like API calls or model inference.
You defined a multimodal pipeline in Python and now need to inspect tables, debug computed columns, roll back changes, and expose HTTP endpoints without writing more application code.
Parse and access specific fields from structured JSON responses returned by language models.
Generate UUIDs for automatic row identification.
Combine data from multiple tables using inner, left, and cross joins.
Undo mistakes, audit changes, and create point-in-time snapshots of your data.
Build reusable aggregation logic for group-by queries and analytics.
An iterator in Pixeltable is a function that expands a single input row into multiple output rows.
Transform a single document, video, image, or audio file into multiple rows for granular processing.
Because typical use cases involve datasets that span multiple time zones, Pixeltable strives to be precise in how it handles time zone arithmetic for datetimes.
Cookbooks
Build the six architectural patterns of agentic AI (prompt chaining, routing, parallelization, tool use, evaluator-optimizer, orchestrator-worker) and the ReAct and planning strategies as Pixeltable tables and computed columns.
Create a retrieval-augmented generation system that answers questions using your documents as context.
Score model output with a second model, as a chain of computed columns
In this tutorial, we'll demonstrate how RAG operations can be implemented in Pixeltable. In particular, we'll develop a RAG application that summarizes a collection of PDF documents and uses ChatGPT to answer questions about them.
In this tutorial, we'll explore Pixeltable's flexible handling of RAG operations on unstructured text.
Create an AI agent that remembers important information across conversations.
Convert a table with computed columns into a callable function for multi-agent workflows and pipeline composition.
Create lookup functions that query tables by key—for customer records, product catalogs, or financial data.
Enable LLMs to call functions and tools, then execute the results automatically.
Cookbooks
Generate concise summaries of long text, articles, or documents using large language models.
Automatically translate content into multiple languages using LLMs.
Identify and extract people, organizations, locations, dates, and other entities from text using LLMs.
Insert a URL as a document, then compare chunking strategies on it
Break PDFs and documents into searchable chunks for retrieval-augmented generation (RAG) pipelines.
Transform office documents into searchable, analyzable text data.
Cookbooks
Run YOLOX object detection in Python on images and video frames, then query and draw the detected objects.
Automatically identify and locate objects in images using YOLOX object detection models.
Understand when to use bounding boxes versus pixel-level masks for image analysis.
Draw bounding boxes on images to visualize object detection results.
Segment every object matching a concept (a short text phrase or a bounding box) using Meta's Segment Anything Model 3.
Automatically create descriptive captions for images using AI vision models.
You have a batch of images that need AI-powered transformations—like turning photos into paintings, adding stylistic effects, or modifying content based on text prompts.
Run the same AI prompt against multiple images automatically, without writing loops or managing API calls.
Use AI vision to extract JSON data from receipts, forms, documents, and other images.
You need to add watermarks to hundreds of different images to protect copyright, add branding, or mark drafts.
You need to make hundreds of images semi-transparent for backgrounds, overlays, or watermarks.
You need to apply filters to hundreds of images—blur, sharpen, edge detection, and other enhancements.
You need to fix inconsistent lighting across hundreds of images—adjusting brightness, contrast, and color saturation.
You need to resize, rotate, crop, or convert hundreds of images—and keep track of all the transformed versions.
You need to convert color images to grayscale for analysis, preprocessing, or model inputs that require single-channel images.
Cookbooks
Transcribe speech with WhisperX in Python, identify speakers, and search the transcript with word-level timestamps.
Convert speech to text locally using OpenAI's open-source Whisper model—no API key needed.
In this tutorial, we'll build an end-to-end workflow for creating and indexing audio transcriptions of video data.
Transcribe audio files and generate summaries automatically using Whisper and LLMs.
Generate YouTube-style chapter timestamps for podcast episodes. WhisperX detects where speech starts and stops, then an LLM summarizes each segment into a chapter title.
Generate natural-sounding audio from text using OpenAI's text-to-speech models.
Extract the audio track from video files for transcription, analysis, or processing.
Cookbooks
Pull frames from video files at specified intervals for analysis, thumbnails, or training data.
Detect objects in a video frame by frame: Pixeltable extracts the frames, runs YOLOX models on them and reassembles the frames into a video. Compare detection models and evaluate them against ground truth.
Automatically find scene cuts, transitions, and fades in video files.
Automatically create preview thumbnails from video files at specific timestamps or intervals.
Burn text, captions, or watermarks directly into video files.
Create videos from text prompts or animate images using Google's Veo model.
Turn a collection of images into an animated video with Ken Burns effects, text overlays, logos, and background music — using a fully declarative Pixeltable pipeline.
Cookbooks
Generate vector embeddings for text data to enable semantic search and similarity matching.
Create a searchable knowledge base that finds content by meaning, not just keywords.
Build visual similarity search to find images that look alike using OpenAI's CLIP model.
Cookbooks
Load data from CSV and Excel files into Pixeltable tables for processing and analysis.
Load structured data from JSON files into Pixeltable tables for processing and analysis.
Load columnar data from Parquet files into Pixeltable tables for processing and analysis.
Load data from Excel spreadsheets (.xlsx) into Pixeltable tables.
Load datasets from Hugging Face Hub into Pixeltable tables for processing with AI models.
Import images, videos, and audio files from S3, GCS, HTTP URLs, or local paths into Pixeltable tables.
When Pixeltable generates media files (thumbnails, extracted frames, processed images), by default it stores them locally.
Convert Pixeltable data to PyTorch DataLoader format for model training.
Send and receive data from PostgreSQL, SQLite, MySQL, TigerData, or Snowflake for use in external applications.
Export Pixeltable tables and queries to CSV, JSON, and Apache Iceberg for spreadsheets, scripts, and the lakehouse.
Create training, validation, and test splits with random or stratified sampling.
Integrations
Pixeltable's Anthropic integration enables you to access Anthropic's Claude LLM via the Anthropic API.
Pixeltable's Bedrock integration enables you to access AWS Bedrock foundation models directly from your tables.
This notebook demonstrates how to use Black Forest Labs (BFL) FLUX models for image generation and editing through Pixeltable.
Pixeltable's Deepseek integration enables you to access Deepseek's LLM via the Deepseek API.
Pixeltable's fal.ai integration enables you to access fal.ai's fast inference models via the fal.ai API.
Pixeltable's Microsoft Fabric integration enables you to access Azure OpenAI models within Microsoft Fabric notebook environments with automatic authentication.
Pixeltable's Fireworks integration enables you to access LLMs hosted on the Fireworks platform.
Use Google Gemini from Pixeltable with a Google AI Studio API key or Vertex AI: generate content, images with Nano Banana and Imagen, and video with Veo.
Pixeltable's Groq integration enables you to access Groq models via the Groq API.
Pixeltable provides seamless integration with Hugging Face datasets and models.
Pixeltable's Jina AI integration enables you to access state-of-the-art embedding and reranker models via the Jina AI API.
This tutorial demonstrates how to use Pixeltable's built-in llama.cpp integration to run local LLMs efficiently.
Pixeltable's Mistral AI integration enables you to access Mistral's LLM and other models via the Mistral AI API.
Pixeltable's Nebius Token Factory integration enables you to access Nebius language and embedding models via an OpenAI-compatible API.
Ollama is a popular platform for local serving of LLMs. In this tutorial, we'll show how to integrate Ollama models into a Pixeltable workflow.
Pixeltable's OpenAI integration enables you to access OpenAI models via the OpenAI API.
Pixeltable's OpenRouter integration enables you to access multiple LLM providers through a unified API via OpenRouter.
Pixeltable's Pydantic integration enables type-safe data insertion using Pydantic models. Instead of inserting raw dictionaries, you can define structured models with validation and insert them directly into Pixeltable tables.
Pixeltable's Replicate integration enables you to access Replicate's models via the Replicate API.
Pixeltable's RunwayML integration lets you generate and transform images and videos using RunwayML's Gen-4.5, Veo, and other models, directly from your Pixeltable tables.
This tutorial demonstrates how to configure Pixeltable to use Tigris for storage. This lets you store unlimited amounts of images in Tigris' global data plane, allowing your images to load fast everywhere.
Use Together AI from Pixeltable for chat completions and image generation.
Twelve Labs provides multimodal embeddings that project text, images, audio, and video into the same semantic space. This enables true cross-modal search - the most powerful feature of this integration.
This tutorial demonstrates how to use Pixeltable's built-in vLLM integration to run local LLMs with high-throughput inference.
Pixeltable's Voyage AI integration enables you to access state-of-the-art embedding and reranker models via the Voyage AI API.
Pixeltable can export data directly from tables and views to the popular Voxel51 frontend, providing a way to visualize and explore image and video datasets.
pxt.create_table() so its cells run without a project. An app puts the same columns on a TableModel in app.py and creates the tables with pxt schema update. The example apps show that form.Install Pixeltable, then take any of these as a starting point.