LiveKit Becomes Voice AI Unicorn

January 27, 2026

🔊 Soundcheck

  • LiveKit hits unicorn status with $100M power-up

  • PersonaPlex: Real‑Time Conversational Voice AI

  • Google settles $68M “false accept” eavesdropping claims

  • VoiceRun Secures $5.5M for Voice AI Platform

Read time: 4 minutes

🔥 Hot Mic

Big moves, deep dives, and standout stories.

LiveKit raised $100M at a $1B valuation, fueling demand for real-time voice AI infrastructure.

LiveKit, originally an open‑source project founded in 2021 by Russ d’Sa and David Zhao, has evolved into a cloud platform powering real‑time voice and video AI. It rapidly gained traction in the voice AI boom, securing OpenAI’s ChatGPT voice mode among its high‑profile customers.

The latest $100 million funding round—led by Index Ventures with returning backers—pushes LiveKit into unicorn territory, reflecting surging enterprise demand for scalable, low‑latency voice infrastructure. The company plans to expand compute, storage, and network services to meet growing global usage.

Key Points:

  • LiveKit raised $100M in new Series C funding.

  • Valuation reached $1B, marking unicorn status.

  • Led by Index Ventures with Altimeter, Hanabi, Redpoint support.

  • Powers OpenAI’s ChatGPT voice mode and other voice AI apps.

Takeaway: LiveKit’s milestone funding and unicorn valuation underscore its role as a foundational infrastructure provider behind the rapid rise of voice‑enabled AI, poised to scale alongside the booming demand for real‑time voice interaction.

Nvidia released PersonaPlex, the first open-source full‑duplex voice model combining seamless dialogue with persona control.

PersonaPlex lets AI listen and respond at the same time, eliminating robotic pauses and creating a much more natural conversational flow. It uses a single 7B‑parameter model rather than separate modules for recognition, language understanding, and speech synthesis. Through hybrid prompting, developers can define both voice and persona with an audio sample and a text prompt, enabling highly customizable characters. Performance benchmarks show PersonaPlex outperforms systems like Gemini Live and Moshi in latency, naturalness scores, voice similarity, and handling interruptions.

Key Points:

  • Full‑duplex design allows listening and speaking at the same time.

  • Hybrid prompting merges voice audio and text role input for dynamic persona control.

  • Zero‑delay switching: just 0.07 s speaker transition latency.

  • Achieved higher naturalness, voice similarity, and interruption handling than rivals.

Takeaway: PersonaPlex marks a major leap in conversational AI by combining real‑time responsiveness with full persona customization in a single open‑source model—no more trade‑offs between naturalness and flexibility.

Google will pay $68 million to settle claims its Assistant mis‑recorded private conversations without consent.

Google has agreed to a $68 million settlement resolving a class-action lawsuit over its Assistant recording conversations without intentional activation. The controversy centered on “false accepts” — instances when devices mistakenly recorded users without the wake phrase. The settlement requires court approval and covers users whose devices triggered unintended recordings since 2016.

Consumers can file claims for up to three Assistant‑enabled devices, with payouts varying based on claim volume. The case echoes privacy suits against other tech giants, notably Apple’s earlier AI‑assistant settlement, underscoring growing scrutiny on voice‑AI consent and security.

Key Points:

  • Settlement addresses unintended “false accept” voice‑assistant recordings.

  • Google denies wrongdoing but opts to settle to avoid litigation.

  • The $68 million fund awaits court approval from Judge Labson Freeman.

  • Claimants with up to three devices may receive variable payments.

Takeaway: This ruling highlights that voice AI needs privacy and consent baked into its design—not just as a feature but as a foundational principle in user trust and compliance.

VoiceRun raised $5.5M to power a transparent, code-first voice AI platform for enterprise-scale deployment.

VoiceRun closed a $5.5 million seed round led by Flybridge Capital, with support from RRE and Link Ventures. The funding is meant to fuel its full‑stack voice AI platform designed for enterprise use.

The company emphasizes a code‑first, developer‑centric model, where businesses retain full control of their application code while VoiceRun manages orchestration of speech‑to‑text, large language models, text‑to‑speech, telephony, and observability. It offers flexible deployment—public cloud, VPC, or on‑premises—to meet compliance and latency needs.

VoiceRun’s tooling includes A/B testing, telemetry, synthetic‑data evaluations, and real‑time improvement workflows. Customers across domains like restaurants, banking, insurance, and telecom have moved from proof‑of‑concept to production in weeks instead of months. The company frames itself as a “voice agent factory” empowering developers and coding agents at scale.

Key Points:

  • Raised $5.5M seed led by Flybridge Capital Partners

  • Enterprise‑grade platform for building and deploying voice agents

  • Code‑first model ensures customer control over application code

  • Supports cloud, VPC, or on‑premises deployment options

Takeaway: VoiceRun is carving out a middle path between slow, bespoke voice AI development and brittle no‑code solutions by offering a fast, flexible, and governed platform that lets enterprises ship scalable voice agents in weeks—without sacrificing control or compliance.

🎙️ Mic Drop

What else is making noise in voice AI.

RingCentral pivots to AI voice, expanding SMB offerings via new AT&T tie-up, aiming to capture a $65B+ market opportunity. (simplywall.st)

Diel-Jerue automates logistics appointment calls with Qued AI voice scheduling, slashing manual labor for fleet operations. (fleetequipmentmag.com)

Just Eat Takeaway.com reveals AI voice assistant for iOS and Android to streamline customer ordering in food delivery. (newsroom.justeattakeaway.com)

Qwen3-TTS debuts as a free, multilingual TTS suite with real-time latency and precise voice control for global developers. (marktechpost.com)

Breez AI lands $1.3M pre-seed to expand its no-code voice orchestration platform for enterprise-scale automation. (waya.media)

India’s Bolna snags $6.3M seed backing to meet surging demand for automated, scalable enterprise voice solutions. (entrepreneurindia.com)

Ringg AI raises $5.5M Series A to enhance its multilingual voice agent platform for global contact centers. (entrepreneur.com)

Enterprises broaden conversational AI use cases beyond CRM, aiming for real-time customer risk detection and smarter forecasting. (forbes.com)

Analysis: SDLC models don’t fit live conversational workflows—agent-centric voice AI product development is recommended. (aijourn.com)

Weathernews debuts SeaNavigator: first conversational AI for maritime captains, offering on-vessel real-time routing and forecasting. (markets.businessinsider.com)

Quick comparison guide to top AI voice agents helping SMEs boost customer service and operational efficiency in 2026. (findarticles.com)

LiveKit picks up $100M Series C to fuel developer-centric real-time voice, video, and spatial media platform expansion. (pymnts.com)

Emerging voice AI orchestration tools are reshaping enterprise AI strategies and application architectures. (venturebeat.com)