Absolutely, you can write custom update options and create an STT node in LiveKit instead of sticking with the default options. LiveKit offers two great ways to customize your experience: • Custom STT Nodes: You have the flexibility to override the

stt_node()

method in a custom

Agent

subclass, giving you full control over the speech-to-text pipeline. This means you can add your own pre-processing, buffering, or even explore alternative STT strategies. For instance, you might preprocess audio frames or dynamically postprocess the transcription text to better suit your needs. • Dynamic Option Updates: Many STT plugins, like Deepgram and Gladia, come with an

update_options()

method. This handy feature lets you tweak STT settings on the fly—things like language, model, punctuation, and custom vocabulary—without having to recreate the STT instance every time. If you'd like, I’d be more than happy to walk you through specific examples or provide more detailed guidance on how to implement these methods. Just let me know! Sources: Pipeline nodes and hooks | LiveKit Docs | livekit-plugins/livekit-plugins-elevenlabs/livekit/plugins/elevenlabs/stt.p... | plugins/deepgram/src/stt.ts