The open-source community is the heart of OpenHome, and today we’re thrilled to spotlight community project Truly Open Source from community member Anand Merneedi!
Open-source technology is rapidly gaining traction in the industry and AI has early adoption of this way of building tech in public. At its heart, open source technology is defined by the free access to a product’s source code, allowing anyone to inspect, modify, and enhance it. This philosophy challenges the traditional model of proprietary software, which restricts access to its source code to protect intellectual property. Open source, by contrast, thrives on the principle of collaboration and transparency, paving the way for a more inclusive and innovative technological landscape.
There is a rich library of open-source models available to integrate with OpenHome. OpenHome’s modularized SDK allows developers to plug-and-play a large selection of open-source models.
About the Project
Anand wanted to see if he could build a version of OpenHome that utilized solely open-source libraries. This means using open-source text-to-speech models, like Whisper or Mozilla DeepSpeech, open-source LLMs like Zephyr, LlaMa, or even a locally-trained model, and open-source voice models like coquiTTS and Tacotron.
He was curious to see how different open-source models could plug-and-play into OpenHome’s SDK, and how each one would perform.
Anand additionally wanted to make all the models run locally, so that you can disconnect your smart speaker from the Internet and still use it.
Demo
About the Developer
This project was made by Anand Merneedi during the OpenHome Voice AI Hackathon. Anand is a developer based in the Bay Area.