Outspeed is a PyTorch-inspired SDK for building real-time AI applications on voice and video input. It offers:
- Low-latency processing of streaming audio and video
- Intuitive API familiar to PyTorch users
- Flexible integration of custom AI models
- Tools for data preprocessing and model deployment
Ideal for developing voice assistants, video analytics, and other real-time AI applications processing audio-visual data.
You can install outspeed SDK from pypi using
pip install outspeed
This would install the core outspeed package.
Read our quickstart guide to get started.
Read the docs to learn more about the SDK.
To deploy your realtime function on Outspeed's infra, you can use the outspeed deploy CLI.
# functions.py contains your realtime function code
outspeed deploy --api-key=<your-api-key> functions.py
Contact us to get an API key and deploy.
Once deployed, you can use the playground in the examples repo to test the deployed code.
All the examples are available in the examples folder.
To install the package so that all examples run, use:
pip install 'outspeed[plugins,torch]'
Or, if you're using poetry:
poetry add 'outspeed[plugins,torch]'
This will install all the additional libraries that are required for examples to work.
To develop on top of the SDK, set environment variable DEV_INFO or DEV_DEBUG to get logs from the SDK for the corresponding log level.
| Feature | Status | Target Release |
|---|---|---|
| Local STT | On the way | Q4 2024 |
| DeepReel Integration (Human Clone) | On the way | Q4 2024 |
| Long Conversation Support | Planned | Q4 2024 |
| Local Model Vision and Text (With Ollama, and vision models) | Planned | Q4 2024 |
| Call Recording | Planned | Q4 2024 |
| Wakeup Word | Planned | Q4 2024 |
| On device models | Planned | Q4 2024 |
| Local TTS | Planned | Q4 2024 |
We welcome contributions to the Outspeed SDK! If you're interested in contributing, please submit a PR or contact us.
