How does the agent handle user interruptions?

It features configurable 'barge-in' strategies, allowing the AI to stop speaking immediately, confirm the interruption semantically, or ignore it based on your specific needs.

Which STT and TTS providers are supported?

The skill supports popular cloud providers like Deepgram and ElevenLabs, as well as local-first options like SherpaOnnx for offline or privacy-sensitive processing.

What audio processing features are included in the pipeline?

The built-in pipeline includes Voice Activity Detection (VAD), Acoustic Echo Cancellation (AEC), RNNoise-based denoising, and automated resampling.

Can I build a voice agent that runs entirely offline?

Yes, by utilizing the SherpaOnnx providers for both STT and TTS along with a local audio backend, you can create fully functional offline voice systems.

Voice Agent Creator

Name: Voice Agent Creator
Author: roomkit-live

byroomkit-live

•

Data Science & ML

Builds production-ready AI voice agents with integrated speech-to-text, text-to-speech, and advanced audio processing pipelines.

The create-voice-agent skill enables Claude to architect and implement sophisticated voice-based AI assistants using the RoomKit framework. It provides a comprehensive toolkit for managing full-duplex audio streams, configuring advanced audio pipelines—including Voice Activity Detection (VAD), Acoustic Echo Cancellation (AEC), and denoising—and integrating leading STT/TTS providers like Deepgram and ElevenLabs. This skill is ideal for developers building responsive voice bots that require low-latency interactions, natural interruption handling (barge-in), and flexible backend support ranging from WebRTC to local hardware.

Key Features

012 GitHub stars

02Standardized event hooks for real-time transcription and session monitoring

03Multi-provider STT/TTS support including Deepgram, ElevenLabs, and SherpaOnnx

04Full-duplex audio pipeline with VAD, AEC, and RNNoise denoising

05Advanced interruption (barge-in) handling with configurable strategies

06Multiple audio backends including FastRTC (WebRTC), SIP, and Local Audio

Use Cases

01Creating voice-controlled assistants with natural conversation flow

02Developing real-time speech-to-text and accessibility tools for live audio

03Building low-latency AI customer service and reception voice bots

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add roomkit-live/roomkit-skills create-voice-agent

For use in Claude.ai and ChatGPT

Key Features

012 GitHub stars

02Standardized event hooks for real-time transcription and session monitoring

03Multi-provider STT/TTS support including Deepgram, ElevenLabs, and SherpaOnnx

04Full-duplex audio pipeline with VAD, AEC, and RNNoise denoising

05Advanced interruption (barge-in) handling with configurable strategies

06Multiple audio backends including FastRTC (WebRTC), SIP, and Local Audio

Use Cases

01Creating voice-controlled assistants with natural conversation flow

02Developing real-time speech-to-text and accessibility tools for live audio

03Building low-latency AI customer service and reception voice bots

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add roomkit-live/roomkit-skills create-voice-agent

For use in Claude.ai and ChatGPT