Skip to main content
If you’ve been wrangling multimodal data with custom Python scripts, DVC for versioning, Airflow for scheduling, and manual processing loops — this guide shows how Pixeltable replaces that plumbing with declarative tables.
Related use case: Data Wrangling for ML

Concept Mapping


Side by Side: Image Processing Pipeline

Process images: generate thumbnails, caption with an LLM, embed for search, version everything.

What Changes


Common Patterns

Video frame extraction

Data versioning

PyTorch export


Next Steps

Data Wrangling for ML

Full use case walkthrough

Extract Video Frames

Frame extraction with FPS control

Export to PyTorch

Convert tables to DataLoaders

Cloud Storage

S3, GCS, Azure, R2, Tigris
Last modified on June 24, 2026