SkyWriter

An Azure Cognitive Services-powered caption generator for OBS

SkyWriter uses Azure Cognitive Services speech-to-text to generate captions and send them to OBS. It will automatically start recognition when it's connected to OBS and detects that a stream is active & your primary microphone source is unmuted. If you mute that source or stop streaming, the session will automatically end.

Requirements

• Windows 10, 1607+ or Windows 11
• OBS 28.0+
• An active Azure Cognitive Services instance

Configuration

Running the app requires a few settings to be configured in appsettings.json:

• AzureSpeechRegion: The Azure region where your Cognitive Services instance. You can find this in your Cognitive Service under Keys and Endpoint. The default is eastus.
• AzureSpeechKey: The key used to connect to your Cognitive Services instance. You can find this in your Cognitive Service under Keys and Endpoint.
• AzureSpeechProfanityLevel: The level for how you want profanity to be handled. Levels and their corresponding numeric values can be found here. The default is 0 (Masked).
• ObsWebsocketHost: The hostname for your OBS Websocket connection. The default is localhost.
• ObsWebsocketPort: The port for your OBS Websocket connection. The default is 4455.
• ObsWebsocketPassword: The password for your OBS Websocket connection.
• ObsMicrophoneSourceName: The name of your primary microphone source in OBS. The default is Microphone.
• SpeechLanguage: The language code for speech to be recognized. Supported language codes can be found here. The default is en-US.
• UseLocalSpeechEngine: true if you want to use the Windows local speech recognition engine instead of Azure. NOTE: This ONLY works on Windows systems, you'll need to setup Voice Recognition in Windows before using this, and it is NOT as good as Azure.