🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Building voice agents? Meet Ink-Whisper: the fastest, most affordable streaming speech-to-text model. 🌎 Optimized for accuracy in real-world conditions 👯 Pair with our Sonic text-to-speech → fastest duo in voice AI 🔌 Plugs into Vapi,Pipecat AI, LiveKit Read more:

Ink-Whisper is fast! Cartesia released a new STT model yesterday, and it's as fast as you'd expect. Streamed transcription finishes in <100ms—before you can say the next sentence. I used it to build a live teleprompter that follows along with what you're saying. Code in 🧵

Inworld's new TTS is rich and expressive! I've been having a lot of fun playing with them. With a competitive cost & low latency TTFB of ~300ms, this is one to watch for in the voice AI space. Give it a try: pip install livekit-agents[inworld]

I think the “shell-qui-hires” (Scale, Windsurf, etc) is one of the worst things to ever happen to startups The point of a startup is there’s assurance of binded fate “shell-qui-hires” undo that bindings As a founder, it’s your job to get everyone home. Full stop.

now you can build voice agents with Spitch thanks to our friends LiveKit.

our design team def dials it to 11.

Next, our end-of-turn model has been significantly improved. In the past, things like punctuation or STT provider formatting could affect the accuracy of the model. The new model is MUCH more resilient, resulting in more accurate turn detection across 14 different languages.

A lot of people have reached out to me for my LiveKit UI integrated with voice agents and avatars, so I made this video to demo it and give the code for free. Free meeting app for voice ai developers to demo their voice agents to clients

Hedra has achieved quite the engineering feat here. They are performing full frame generation - every pixel is synthesized, not just the lips area. Being able to do it at $0.05/min is just 🤯🤯🤯

Since Hedra allows you to create avatars with ANY image, you can create new avatars in real time by generating new images with an LLM. In this example, we're using GPT-4.1-mini to generate a random avatar. We pass that image to the avatar session, and Hedra will animate it.

I don't think people understand that raising money is like, one of the easiest parts of starting a company You're pitching to sell 100% of the next years of your life If your pitch makes sense, they *will* sign you up for that work Be very deliberate about what you do

since flutter-webrtc is relied on by many products, its main functions have been implemented and become stable, has more than 50k+ downloads per month recently, I think it is time to update its version number to 1.0. github.com/flutter-webrtc…

The moment things changed for us? Switching to LiveKit LiveKit handles the entire real-time media pipeline: •TTS •STT •LLM integration •Telephony •And most importantly… turn detection 🧠

I built a simple voice assistant in 70 lines of Python code. It uses: • LiveKit - The voice agent • AssemblyAI - To turn your voice into text • OpenAI - The brain of the agent, and to turn text into audio There's something really cool about this: