hotwirednews Get GrokBot

Week 8

16–22 Feb 20262026-W08 · 25 filings · 18 companies · 10 net-new

The Orchestrate layer led the week with 7 of 25 filings, from Daily / Pipecat, ElevenLabs, Momentum and BHASHINI.

Hearspeech-to-text2

Gnani.aiModel providermodel release2026-02-19net-new

Gnani.ai ships Vachana STT and TTS under Inya VoiceOS

Sovereign Indic STT/TTS with a claimed ~10M calls/day and sub-5% Hindi WER — all Gnani-stated, but telephony-first and on-prem options matter for Indian BFSI/gov.

Also Speak

Details1 source

Gnani.ai launched Vachana STT and Vachana TTS under its Inya VoiceOS stack at the India AI Impact Summit: India-hosted speech models for Indic languages with claimed sub-5% Hindi/Indian-English WER and zero-shot voice cloning under 10 seconds.

Vachana STT + TTS

DeepgramModel providermodel release2026-02-19

Deepgram adds Arabic across 17 dialects to Nova-3

Dialect coverage makes a production STT model usable across Arabic-speaking markets without selecting a separate model for each regional variety.

Details2 sources · 1 X post

Deepgram made Nova-3 Arabic speech-to-text available across 17 dialects for call-centre, voice-agent, and transcription workflows.

Nova-3 Arabic

‼️ Demo video below: Arabic is now available on Nova-3 — across 17 dialects. Nova-3 is built to capture real-world Arabic as it’s actually spoken around the world. That means higher accuracy, better understanding of regional nuance, and more reliable Voice AI experiences. Whether you're building call center automation, AI voice agents, or transcription workflows, you can now deploy Arabic with confidence. 🎥 Watch the demo to hear it in action. 🚀 Try it yourself in the Playground: https://t.co/twpVGFPnTl #VoiceAI #SpeechToText #Arabic #AI #Developers

https://x.com/DeepgramAI/status/2024581379762298987

Speaktext-to-speech3

Daily / PipecatPlatformlaunch2026-02-20

Pipecat 0.0.103 adds Deepgram SageMaker TTS and Rime Arcana defaults

Self-hosted Deepgram TTS on SageMaker matches the existing SageMaker STT path.

Details3 sources

pipecat-ai 0.0.103 adds DeepgramSageMakerTTSService for Deepgram TTS on SageMaker HTTP/2, defaults RimeTTSService to model arcana, and adds Arcana/Mistv2 InputParams.

pipecat-ai 0.0.103

KittenMLModel providermodel release2026-02-19net-new

KittenML ships KittenTTS 0.8 nano, micro, and mini CPU TTS models

Sub-100 MB English TTS under Apache-2.0 sits at the opposite end from cloud SKUs; nano int8 shows about 195,000 Hub downloads at capture, and KittenTTS 2 (2026-10-02-kittenml-kitten-tts-2) keeps these eight voices while moving to a revenue-capped licence.

Details4 sources

KittenML released KittenTTS 0.8, Apache-2.0 open TTS weights in nano, micro, and mini sizes with eight built-in voices that run on CPU without a GPU.

KittenTTS 0.8

ArgmaxPlatformlaunch2026-02-19net-new

Argmax open-sources TTSKit for on-device speech on Mac and iPhone

An open on-device TTS path brings a voice model runtime to Apple hardware without requiring a hosted speech API.

Details2 sources · 1 X post

Argmax open-sourced TTSKit for running Qwen3-TTS on Mac and iPhone, with faster-than-real-time audio and sub-200 ms time to first byte.

TTSKit

We are open-sourcing TTSKit! Run state-of-the-art text-to-speech models on your Mac and iPhone. The launch version supports @Alibaba_Qwen Qwen3-TTS and generates audio faster than real-time playback with sub-200 ms time-to-first-byte. Voice cloning and advanced speed optimizations will be in the next version. Link to the GitHub repo and models on @huggingface in comments.

https://x.com/argmax/status/2024520937203454405

Showavatars, video4

AnamPlatformpartnership2026-02-19net-new

Anam launches a no-code Framer integration for embedded avatars

A no-code site-builder integration lowers the implementation cost for putting a conversational avatar on a public web page.

Also Orchestrate

Details2 sources · 1 X post

Anam's Framer plugin lets users embed a real-time Anam avatar in a Framer site using code generated from Anam Lab.

Framer x Anam integration · vendor Framer

Today, we’re launching our official Framer x Anam integration. It’s a no-code approach where users can design their site in @framer and add an avatar in minutes. https://t.co/Rp3DXzybtT

https://x.com/Anam__ai/status/2026737778344485118
TavusModel providermodel release2026-02-18

Tavus introduces Phoenix-4 for real-time human rendering

Full-frame runtime rendering and conversational responsiveness make the avatar renderer part of the live-agent loop rather than a post-produced video step.

Details2 sources · 1 X post

Tavus introduced Phoenix-4 as a real-time human-rendering model for conversational avatars.

Phoenix-4

Introducing Phoenix-4, the most advanced real-time human rendering model ever built. Its the first real-time model to render every pixel at runtime, generate and control emotional states, listen actively, and actually behave the way humans do in conversation. https://t.co/IP4hZowliA

https://x.com/tavus/status/2024163626765148488
LiveKitPlatformlaunch2026-02-17

LiveKit Agents 1.4.2 adds TruGen avatar plugin

Avatar additions for 1.4.2.

Details1 source

livekit-agents 1.4.2 adds a TruGen avatar plugin and documents LemonSlice under video avatars.

livekit-agents 1.4.2

AnamPlatformmodel release2026-02-17net-new

Anam introduces cara-3, a real-time avatar model

A new avatar model gives real-time agents a model-level rendering option; the comparative evaluation language remains Anam's claim.

Details2 sources · 1 X post

Anam introduced cara-3 as a real-time avatar model that generates photorealistic AI personas for live interaction.

cara-3

Introducing cara-3 - the highest-rated real-time avatar model ever tested. It generates photorealistic AI personas in real time. Independent evals: #1 across every metric. Experience the face of AI → https://t.co/vgzfMVGu9x 🧵

https://x.com/Anam__ai/status/2023767843876344247

ThinkLLMs, speech-to-speech3

KrispPlatformAPI2026-02-18net-new

Krisp launches Real-Time Voice Translation SDK for CX platforms

CX vendors get in-call speech-to-speech translation as an SDK on 18 February — production-proven in Krisp's own contact-centre product, not a greenfield demo.

Also HearSpeakConnect

Details2 sources

Krisp released a Real-Time Voice Translation SDK so customer-experience platforms can embed speech-to-speech translation in live calls across 60+ languages, with local noise cancellation and custom vocabulary.

Voice Translation SDK

FlashLabsPlatformlaunch2026-02-18

FlashLabs launches FlashAI 2.0 integrated voice-agent platform

Another vertically integrated voice-agent stack (Chroma) shipping 18 February against cascaded APIs — early-access only, so treat production footprint as unproven from the wire alone.

Also OrchestrateConnect

Details1 source

FlashLabs announced FlashAI 2.0, a vertically integrated enterprise voice AI platform on its Chroma speech stack for contact-centre and sales agents, with human-in-the-loop escalation and an early-access programme for the first 100 companies.

FlashAI 2.0

Gnani.aiModel providermodel release2026-02-17net-new

Gnani.ai unveils Inya VoiceOS 5B speech-to-speech research preview

Sovereign Indic S2S research preview at 5B parameters with a stated 14B follow-on — press-sourced vendor figures, not a dated public API GA.

Also SpeakHear

Details2 sources

Gnani.ai unveiled Inya VoiceOS at the India AI Impact Summit: a 5-billion-parameter speech-to-speech foundational model under the IndiaAI Mission, released as a research preview ahead of a planned 14B system.

Inya VoiceOS

Orchestrateframeworks, evals7

Daily / PipecatPlatformAPI2026-02-22

Pipecat Context Hub 0.0.3 adds search_api over an AST index of Pipecat source

Agents can read framework signatures from the index instead of opening installed source, though the 5,075 chunks cover only src/pipecat/ at this release; other repositories follow in 0.0.6 (2026-03-03-pipecat-context-hub-0-0-6).

Details1 source

Pipecat Context Hub 0.0.3 adds a search_api MCP tool that queries class, method, and function definitions extracted from the Pipecat framework source.

Pipecat Context Hub 0.0.3

Daily / PipecatPlatformlaunch2026-02-20

Pipecat Context Hub 0.0.1 releases a local MCP server for Pipecat docs and examples

This is the first row for the hub and it shipped the same day as framework 0.0.103 (2026-02-20-pipecat-0-0-103-daily-sip-transfer); the hub indexes public repositories and is not part of the framework package.

Details2 sources

Pipecat Context Hub 0.0.1 gives Claude Code, Cursor, VS Code, and Zed five MCP retrieval tools over Pipecat documentation, framework code, and examples, running locally with no API key.

Pipecat Context Hub 0.0.1

Daily / PipecatPlatformAPI2026-02-20

Pipecat 0.0.103 makes user idle timeout updatable at runtime

Bots can raise or clear idle prompts mid-call without rebuilding the aggregator.

Details3 sources

pipecat-ai 0.0.103 adds UserIdleTimeoutUpdateFrame and always constructs UserIdleController (timeout 0 disables).

pipecat-ai 0.0.103

ElevenLabsPlatformlaunch2026-02-19

ElevenLabs adds controlled A/B testing to ElevenAgents

A/B testing moves agent tuning from anecdotal call review to controlled comparisons on real interactions, while keeping configuration changes auditable.

Also SpeakUse

Details2 sources · 1 X post

Experiments lets teams expose controlled agent-configuration variants to live traffic and compare operational and business outcomes.

Experiments in ElevenAgents

Introducing Experiments in ElevenAgents - the most data-driven way to improve real-world agent performance. Experiments enables you to run controlled A/B tests to measure which agent configuration works best - from prompt structure to workflow logic, voice, and personality.

https://x.com/ElevenLabs/status/2024539644579271048
MomentumPlatformacquisition2026-02-18net-new

Salesforce agrees to acquire Momentum's conversational-insights platform

The acquisition gives Salesforce a direct path from recorded voice and video conversations to agentic actions, complementing live voice handling with post-conversation context and revenue orchestration.

Also HearUse

Details1 source

Salesforce agreed to acquire Momentum, whose ingestion engine captures and structures customer conversations from third-party video channels for Agentforce and Slackbot workflows.

Momentum acquisition by Salesforce

BHASHINIPlatformlaunch2026-02-18net-new

BHASHINI launches VoicERA, an open-source voice-agent stack for Indian government telephone services

India's language mission chose an open Pipecat-based runtime with swappable carriers rather than a procured platform, so state voice services can be built without one vendor; DARPG's Samadhan Didi grievance chatbot (2026-05-30-darpg-samadhan-didi-voice-chatbot, W22) later used BHASHINI speech for a central-government service, though its release does not say it runs on VoicERA.

Also Connect

Details6 sources · 1 X post

A self-hosted, Apache 2.0 stack for building Indian-language phone agents, deployed on BHASHINI's national language infrastructure so ministries and states can stand up voice services without a single vendor.

VoicERA · vendor Pipecat · delivered by EkStep Foundation

VoicERA expands BHASHINI's vision beyond translation into: Real-time speech systems, Conversational AI, Multilingual telephony at population scale

https://x.com/_BHASHINI/status/2024067083319849426
ElevenLabsPlatformlaunch2026-02-17

ElevenLabs launches ElevenAgents for Support for no-code customer-service workflows

The support-specific packaging connects agent configuration to existing SOPs and help-centre content, shortening the path from documentation to a deployed voice or digital workflow.

Also SpeakUse

Details2 sources · 1 X post

ElevenAgents for Support turns support documentation into production-ready conversational agents across chat, phone, email, and messaging.

ElevenAgents for Support

Introducing ElevenAgents for Support - a dedicated offering to help customer support teams transform manual processes into production-ready agent workflows.

https://x.com/ElevenLabs/status/2023816863239508477

Connecttelephony, channels1

Daily / PipecatPlatformAPI2026-02-20

Pipecat 0.0.103 adds Daily SIP transfer and REFER frames

Warm/cold SIP transfers become frame-driven instead of out-of-band Daily API calls.

Details3 sources

pipecat-ai 0.0.103 adds DailySIPTransferFrame and DailySIPReferFrame so pipelines can queue SIP transfer and REFER on Daily transport.

pipecat-ai 0.0.103

Useagents in market5

Sarvam AIPlatformpartnership2026-02-20

Sarvam and Maharashtra sign an LoI for sovereign AI infrastructure and public-service voice work

The agreement links Indian-language voice and speech capability to state-controlled infrastructure, while the public posts do not identify a deployed voice service or launch date.

Also SpeakConnect

Details3 sources · 2 X posts

Sarvam and the Government of Maharashtra agreed to build sovereign AI infrastructure for state deployments, with the partnership later describing work across citizen outreach, voice, and speech technologies.

Maharashtra sovereign AI partnership

We're excited to partner with the Government of Maharashtra. Through this partnership, we aim to strengthen the state’s sovereign compute capacity and deliver tangible impact across healthcare, education, agriculture, and governance.

https://x.com/SarvamAI/status/2027326082077327706

Announcing our partnership with the Government of Maharashtra. Sarvam has partnered with the Government of Maharashtra to bring Indus into the daily work of 2,500 officials across the state. Officials will use Indus to research policies and precedents, draft notes and replies, translate and summarise documents, analyse data, and work with scanned records across 22 languages, including Marathi. The deployment will run on secure, audited infrastructure located entirely within India, keeping government data within the country and on systems the state can inspect. We’re looking forward to building on this work with the Government of Maharashtra across citizen outreach, voice, and speech technologies for public services. Read more: https://t.co/062KEmf3gF

https://x.com/SarvamAI/status/2087533012091379715
Simple AIPlatformpartnership2026-02-19

Simple AI partners with Cartesia on enterprise sales voice agents

This is a named production deployment rather than a model listing: the voice supplier ties its latency and voice stack to a sales-agent operator's live call volume.

Also SpeakOrchestrate

Details2 sources · 1 X post

Simple AI partnered with Cartesia to power enterprise sales agents with low-latency voices, and says the agents have handled millions of production calls for brands including DoorDash and xAI.

Cartesia-powered enterprise sales voice agents · vendor Cartesia

Sales voice agents fail if they sound robotic 🤖❌ That’s why @simpleailab partnered with Cartesia to power enterprise sales agents with the highest-quality voices and fastest time-to-first-token on the market. The result? ✅30% higher conversion/upsell rates than human reps ✅100x scalability during peak traffic ✅Millions of production calls for brands like DoorDash & xAI Read more on how @catheryn_li and the Simple AI team is building the future of autonomous sales with Cartesia ⬇️

https://x.com/cartesia/status/2024624529826512967
Fortuna HealthPlatformlaunch2026-02-19net-new

Fortuna Health adds a voice agent to its Medicaid renewal platform for reminder calls and inbound support

Deliberate background office noise on calls about public benefits raises a disclosure question that the post does not address; Highmark Wholecare's Pennsylvania expansion (2026-10-07-highmark-wholecare-fortuna-medicaid-voice, W41) is the first named plan to add the agent, about eight months later.

Details2 sources

Fortuna's agent calls members about upcoming renewals, documents, and deadlines, and answers inbound calls with status updates and common answers before routing complex cases to navigators or the state or county.

Fortuna Voice AI

TravelersEnterpriselaunch2026-02-18net-new

Travelers launches AI Claim Assistant voice agent for auto damage claims on OpenAI Realtime

A Dow Jones component insurer with nearly $49 billion of 2025 revenue runs first notice of loss on OpenAI's Realtime API; the 85–90% completion figure comes from OpenAI's 2 June case study, not from Travelers.

Also Think

Details5 sources

Travelers began handling auto damage claim calls with AI Claim Assistant, a fully agentic voice service built on OpenAI models and the Realtime API that takes callers from policy questions through claim submission.

AI Claim Assistant · vendor OpenAI

PolyAIPlatformfunding2026-02-17net-new

PolyAI raises $200M for its voice-agent platform

The round gives a large voice-agent operator capital while its stated footprint is measured in deployed calls and named restaurant and hospitality customers.

Also OrchestrateConnect

Details2 sources · 1 X post

PolyAI announced that it raised $200M from Nvidia, Khosla Ventures, and other venture investors.

$200M funding round

PolyAI has raised $200M from Nvidia, Khosla Ventures, and multiple top VCs. We're one of the fastest-growing companies in the UK, and we handle 500M+ calls for: • Marriott • PG&E • Gordon Ramsay's restaurants • And 3,000 more real deployments Which means that if you've ever called them, chances are you've talked to our voice agents. Every restaurant we onboard books thousands in revenue within 30 days. But how? Because PolyAI works 24/7, answering every call in <2 seconds, and we also: • switch between 45+ languages • handle payments & cancellations • verify identities • and even upsell your services If you want to try creating an agent with PolyAI, we built Agent Studio Lite to make it easy. Just enter any URL, and in 5 minutes it will analyze your website and build a working agent. We're opening early access to a limited number of people. Comment "PolyAI" and we'll add you to the waitlist and give you 3 months for free!

https://x.com/polyaivoice/status/2023789465509015972