Hey LiveKit community! I've been working on a project for the last month that was built entirely on top of LiveKit, and I wanted to share it back with the community. It's an open-source platform called InputRight. It lets you deploy an AI voice agent that turns a spoken conversation into 100% accurate, user-verified data. The core of the idea is a "voice-to-verified-form" workflow that I think is a pretty cool use case for the platform. I honestly could not have built this without the

livekit-agents

SDK for Python. The framework made it incredibly simple to stream audio into our AI pipeline, and the two-way RPC functionality was the absolute key to making the real-time, human-in-the-loop form verification possible. ⭐ GitHub Repo (Apache 2.0 License): https://github.com/jeffo777/input-right 🎬 Here's a 2-minute demo video showing it in action: The open-source kit is a complete, runnable monorepo with a React/Next.js frontend, a FastAPI token server, and of course, the Python agent. I'd be honored if you'd check it out, and any feedback, questions, or GitHub stars would be hugely appreciated! Thanks to the whole LiveKit team for building such an incredible set of tools.

I built this with livekit audio: https://aiphonebooth.xyz

👋 Hey everyone, I’m building DataLoom — it lets you connects data from various sources and helps uncover insights + take action all in one place. I am planning to include

livekit

conversation logs as a data source. For example, if you’re using

LiveKit

, you could: • Track weekly call drop rates automatically • See which sessions had the most engaged participants • Get a daily summary of call quality trends in Slack • Your use case here ... I’d love to take care of the boring analytics work for free while I’m building this, and learn more about your workflow along the way. If this sounds useful, feel free to DM me or schedule a 1:1 call using my website.

Hi everyone, breaking the pattern here, I haven’t built an AI voice agent with LiveKit. Instead, together with a friend we built Hopp, an open source remote pair programming app. Feel free to check out our implementation. I also want to give a shoutout to the LiveKit team, thanks for the great product 🚀.

Hey LiveKit community! 👋 We just launched Macrowave on Product Hunt today - a P2P radio station app for macOS/iOS that we built using LiveKit under the hood! The technical challenge: We wanted to create a way to broadcast any audio source from your Mac (Spotify, YouTube, DJ software, vinyl via audio interface) with ultra-low latency so friends could experience music together in real-time - like sharing the aux cord over the internet. How LiveKit made it possible: • Near-zero latency broadcasting (< 200ms) - crucial for synchronized listening experiences • Seamless WebRTC handling for our P2P architecture • Rock-solid audio quality that preserves the original source • Scales beautifully from intimate 5-person sessions to larger broadcasts The fun part: We wrapped it all in a skeuomorphic radio interface that makes broadcasting as simple as turning a dial. Listeners don't even need the app - just a web link! Since soft launch, we've seen DJs stream live sets, vinyl collectors host listening parties, and remote teams create shared "office radio" while working. The LiveKit infrastructure has been absolutely bulletproof. 🚀 Would love your feedback and technical questions! Check it out: [Product Hunt link] Building with LiveKit has been incredible - thanks for creating such amazing infrastructure that lets us focus on the user experience!

Hi LiveKit community, I just published an open source project to do post call analysis for voice agent calls. Here is the link to the repo - https://github.com/DrDroidLab/voicesummary. Currently have added pause, interruption and other agent behaviour markers using Librosa. Looking to build more qualitative markers as well like deviation from script etc so that builders know where to improve their agent prompts. Looking for some early feedback. Thanks

Hey Livekit community, you might be aware of the open source observability Whispey for livekit agents. added a new Whispey feature: give real-time voice feedback to voice AI agents during conversations without impacting the flow. This lets clients share feedback instantly, and devs can see it mapped to each turn with full trace details. See how we do it here - https://youtube.com/shorts/wIKw6HoWxjo?si=J_dScaqjHEJDwfZ8 whispey.xyz. github - https://github.com/PYPE-AI-MAIN/whispey

Shayne demonstrates using GPT-5 with LiveKit Agents https://x.com/shayneparlo/status/1961151611285217479

set the channel topic: Share what you have been building with LiveKit

Hey Everyone! 👋 I’m excited to share a quick demo of AceInt - Your Personal AI Interviewer — a real-time voice agent with an avatar, built using LiveKit! This real-time voice agent can talk, listen, understand, and provide a full interview experience. It’s designed to help users: • Boost confidence • Improve communication and articulation • Practice technical, coding, HR https://aceint.ai/ Here’s a sneak peek of what I’ve been building — your own personal AI interviewer for realistic, interactive practice! Product Demo 1: Product Demo 2 (with Avatar): https://www.youtube.com/shorts/SU8O78pyh_8 Would love to hear your thoughts and feedback. 🚀

I tried out LiveKit's new Avatar provider plugin Anam, and it's really good!

hey hey, livekit and bluejay are hosting an event together on building voice agents the right way, come if ur interested! 🙂 https://luma.com/4g4oocpa

hey everyone! we just launched our hardware - a voice first, screen-free curiosity device for kids & announced our fundraise! we're powered by livekit! https://x.com/theBhulawat/status/1966200231705595932

Hey everyone! Just launched Vaani, a zero code voice AI solution. With Vaani, you can build voice agents ready for production just by editing simple JSON files. You don't have to hardcode every interaction anymore. For Developers & Open Source Fans: Vaani is fully open source and designed with your needs in mind. - Pluggable & Powerful: Built on LiveKit Agents SDK with OpenAI, Deepgram, ElevenLabs, Sarvam, Pinecone integrations. - Batteries Included: Comes with RAG, conversation memory, and a tool system out of the box. - Works Everywhere: macOS, Linux, Windows, Cloud, Docker. Give it a star on GitHub & help us grow: https://github.com/namish800/vaani For Startups, Agencies & Enterprises: No need to start from scratch. Vaani can power your customer service bots, sales automation, or appointment schedulers. Our enterprise platform adds: • Real-time analytics & monitoring • MCP server support • Professional support & custom solutions • Embeddable & Customizable widgets Learn more about the platform: https://vaani.greatapeai.com How You Can Get Involved: • Developers: We're looking for contributors, especially for OpenAPl tool integration & MCP support! • Businesses: We're seeking beta testers for our enterprise platform and partners for custom solutions. We're opening our beta program to just 50 early adopters. If

hey everyone did some testing, can you guys please look at it and validate it ? https://medium.com/@adarshsingh09.09.03/end-of-utterance-detection-in-voice-agents-why-models-arent-the-latency-culprit-379f165ed23d

Hey Folks, we have published a blog to connect Livekit to Whatsapp Calling. Let me know if you face any troubles in setting it up. https://www.linkedin.com/feed/update/urn:li:activity:7373307546050179072/

hey everyone How are you evaluating your voice AI agents? https://www.linkedin.com/posts/deepesh-agrawal-238601198_voiceai-healthcareai-healtht[…]m=member_desktop&amp;rcm=ACoAAC5dEAwBJKjC-BsatptewUZdolcYq3cHrcE

Working in the Enterprise space all day long, sometimes it's nice to simply build projects for fun 😁. An interview with Russ d'Sa in which he discusses KITT in Knight Rider, inspired this sci-fi themed demo - The Weyland Museum Curator. 👽 Although this is demo is a bit cheeky, it was a great exercise in showcasing how easy it is to publish and subscribe to tracks with the LiveKit SDK. Try doing this with bare-bones WebRTC... forget about it! There are probably much better ways to do this and many more optimizations to be had here, but this simple demo showcases the following: • Livekit agents SDK with Deepgram, OpenAI, and Eleven labs pipeline • Hedra's live avatar service starring David via hedra plugin. BTW Hedra has done some incredible work here! Go check them out! • Various audio tracks being played using the BackgroundAudio player • Tool calls for exhibit context and slides • The slides are being published using the VideoFrame and track • Starter templates from Hedra, Python agent backend, Playground frontend app • There was is fun logic in there in getting David to accept correct passwords and deny anything outside of the correct answers.

You may enjoy this chat with DZ about building Voice AI: https://www.linkedin.com/events/7369757397759823873/

Building specialized STT, 30% better WER from any other provider, average of 350ms, compatible Livekit https://latice.ai/

I wrote an lk-gui through AI programming. lk-gui is a GUI application for LiveKit CLI, providing an intuitive graphical interface to simplify LiveKit command-line operations. https://github.com/zesun96/lk-gui

I had a couple of requests for the repo to my Hedra Demo above, so sharing it below. I couldn't edit the original post. Keep in mind it's very much unpolished demo composed from various templates 😁. But, feel free to use it as a reference. LiveKit & Hedra - Alien Avatar Museum Curator Demo

It is fun to see how folks are using real-time AI with LiveKit

Hello from the UK 🇬🇧 We're building an AI financial planner - helping people navigate complex money decisions 💰 Built on top of LiveKit, check it out! https://www.linkedin.com/posts/dimatarasenko_today-im-so-excited-to-introduce-warren-[…]m=member_desktop&amp;rcm=ACoAABzQeEIBS6niQh1SxZaVcpjW1iIm7cbydY8

Hey gang, we open sourced our on-device oriented TTS model that you can play around with here: https://huggingface.co/spaces/neuphonic/neutts-air Checkpoint and more deets here: https://huggingface.co/neuphonic/neutts-air

This is a very cool ("fast") app:

so I got tired of calling myself over and over again or calling random IVRs to test IVR traversal I made a tool that allows you to make test IVRs and call them via real numbers ( sip trunking coming soon ) to test your agent you can check it out here https://testvoice.ai/ Let me know if you have any feedback.