Squawk - Real-Time Local Text-to-Speech with AI v0.0.4

The OBS Squawk plugin adds powerful voice generation capabilities to OBS by leveraging sherpa-onnx. With this plugin, you can generate speech on the fly and in real-time inside OBS without any external services or access to the network.

If you like this work drop a review here or the Github repo, support our work on Github, Patreon or OpenCollective. Learn about AI for streaming and content on OCC AI.
Need help? Live support on https://discord.gg/Z6CuJQH4S2

Features​

• OBS Audio Source: Seamlessly integrates with OBS as an audio source.
• Sherpa-onnx: Utilizes sherpa-onnx for high-quality voice synthesis and cloning. Everything built-in - not relying on any external software!
• Cross-Platform: Works the same way on any operating system that OBS suports: Windows, Mac and Linux.
• Extensive Voice Library: Access to a huge library of pre-trained voices for dozens of languages.
• Automated Generation: Monitor an OBS source or text file, when the content changes speech will be generated. Perfect for any automation or for live Transcription-Translation-Generation use case.
• Real-Time & Lightweight: The very efficient VITS architecture for speech generation is extremely resource efficient and runs real-time on a modest CPU.

Usage​

1. Open OBS and add a new Source: Select the "Squawk Text to Speech" from the list of available audio sources.
2. Configure the source settings:
   • Select Voice: Choose a pre-trained voice package from the library. Models will be downloaded as necessary.
   • Select the Speaker ID: some voice packages have multiple (even 100s) speakers.
   • You can generate speech directly from the plugin settings by clicking the button.
   • Set up the monitoring of a Text source or a file.
3. Send text to the monitored text source or text file to produce the speech audio.