An open source container for transformative AI.

All Python based. Designed for applications using interchangeable models that ingest and output real-time video.
Getting Started






Tensor Threads are distributed processing nodes each capable of supporting hundreds of AI inferences with zero switching delay, allowing for multistage model processing.

Integrated inside the Ooblex container means tensor threads support both high bandwidth video in and coordinated video out.

Tensorflow, PyTorch, and Caffe are supported.



Ooblex is for the data scientist whose ready to push into real-time video processing with an easy to use python based container. Test Ooblex out using our built in video transformation application.

Ooblex is open sourced under the GPL 3.0 open source license. It is with everyones help that we will build the container for real-time AI.

Join us in building that future.


IBM is a part of our community.

Using Ooblex, IBM clients are discovering new possibilities in facial recognition and fraud detection use cases.
Case Study