Skip to main content
The AssemblyAI Self-Hosted Streaming Solution lets you run our real-time transcription service inside your own infrastructure. Audio, transcripts, and PII never leave your network. The only outbound traffic is license validation and, for usage-based contracts, usage metadata sent to AssemblyAI.
Self-hosted streaming requires an upfront commercial commitment of $50,000. Contact our sales team to discuss your needs and learn more about our self-hosted offering.
Everything you need to deploy and operate the stack lives in the public streaming-self-hosting-stack repository, and that repository is the source of truth. This page explains what the offering is and where to find things. It deliberately does not repeat hardware requirements, setup steps, or configuration details, because those change with each release.

​
How it works

The stack is a set of Docker Compose services. A WebSocket gateway accepts client connections, one or more GPU-backed ASR containers run the models, and a license proxy validates your license and reports usage. Clients connect over the same WebSocket protocol as the hosted Streaming API, pointed at your own endpoint, and choose the model per session.

​
What the repository contains

  • Supported models and GPU requirements. The top-level README lists which models you can serve and the GPU each one needs.
  • Prerequisites. License file, Docker, NVIDIA Container Toolkit, and credentials for pulling images.
  • Streaming stack. Compose files, environment reference, an example client, and production sizing and deployment recommendations live in the repository’s streaming directory.
  • Serverless option. Each stack can also be deployed to serverless GPUs on Modal. The same directory documents it.
  • Changelog. Model improvements, new API features, and breaking changes are recorded in the README changelog, and tagged versions appear on the Releases page.

​
Getting started

  1. Ask your AssemblyAI representative for a license file and image-pull credentials.
  2. Complete the prerequisites in the repository README.
  3. Follow the streaming stack’s README to configure, start, and test the stack.

​
Staying up to date

New releases can change the models served, the GPU requirements, and the speech_model values clients pass. Review the changelog before upgrading and watch the Releases page for breaking changes.

​
Support

For deployment questions, image access, license issues, or to report bugs, contact your AssemblyAI representative.