hotwirednews Get GrokBot

Week 36

31 Aug – 6 Sep 20262026-W36 · 33 filings · 25 companies · 7 net-new

The Orchestrate layer led the week with 7 of 33 filings, from LiveKit, SoundHound AI, Wonderful, Bolna and 2 more.

Hearspeech-to-text6

LiveKitPlatformlaunch2026-09-05

LiveKit Agents 1.8.0 adds AssemblyAI universal-3-6-pro and Deepgram Flux STT options

STT model and option wave beside the OTel break.

Details1 source

livekit-agents 1.8.0 adds AssemblyAI universal-3-6-pro streaming, Deepgram Flux numerals/profanity/redact options, ElevenLabs Scribe v2 secondary_languages, and Google chirp_3 custom prompts.

livekit-agents 1.8.0

MicrosoftModel providermodel release2026-09-03

MAI-Transcribe-2: multilingual ASR and diarization at $0.10/hr

$0.10/hr launch price with diarization and timestamps. Microsoft frames it as limited-time through year-end. Treat “fastest/cheapest in the world” as marketing.

Details2 sources · 1 X post

Microsoft launched MAI-Transcribe-2, a multilingual ASR model with diarization, timestamps, keyword biasing, and launch pricing of $0.10 per hour of audio.

MAI-Transcribe-2

MAI-Transcribe-2: the highest quality, cheapest transcription at the fastest speed! 10x faster that GPT-Transcribe. Now available on Microsoft Foundry.

https://x.com/MicrosoftAI/status/2095521860184363074
MicrosoftModel providermodel release2026-09-02

Microsoft Research releases VibeVoice-ASR-Streaming open weights for streaming speaker-attributed ASR

Self-hosting builders get open, MIT-licensed streaming diarised ASR a month before Microsoft AI's hosted MAI-Transcribe-2-Streaming (2026-10-01-microsoft-mai-transcribe-2-streaming); the two come from different Microsoft groups and neither source links them.

Details4 sources

Microsoft Research published MIT-licensed 1.5B and 7B weights and inference code for one model that transcribes who said what as audio arrives, with hotword biasing across 10 languages.

VibeVoice-ASR-Streaming

MoondreamPlatformlaunch2026-09-01net-new

Moondream Photon 2.1 adds streaming speech recognition for Whisper, Qwen3-ASR, and Parakeet

Throughput gains are vendor-run against each model's reference engine; Photon became the runtime for Moondream's own Parakeet Redux and Ultra weights three weeks later, on 22 September.

Details2 sources · 1 X post

Moondream's Photon inference engine added file and live-stream ASR for Whisper large-v3-turbo, Qwen3-ASR 0.6B and 1.7B, and Parakeet TDT 0.6B v3, alongside NVIDIA B200 support across its model catalogue.

Photon 2.1

MetaModel providerlaunch2026-09-01net-new

Meta Muse Voice Transcribe: streaming ASR plus 20+ speaker diarization

Joint streaming ASR, diarization, and endpointing in 80ms chunks. Vendor: #1 on Artificial Analysis streaming STT as of 1 Sep.

Details1 source · 1 X post

Meta released Muse Voice Transcribe, a streaming ASR model that jointly does transcription, 20+ speaker diarization, and endpointing with multilingual code-switching.

Muse Voice Transcribe

Introducing Muse Voice Transcribe, the first real-time audio perception model from Meta Superintelligence Labs.

https://x.com/AIatMeta/status/2094839236016976028
SonioxModel providerAPI2026-08-31

Soniox opens an India region for in-country speech-to-text and text-to-speech

Indian BFSI and healthcare voice-agent builders get in-country STT and TTS from one vendor, two weeks before Deepgram's India endpoint (2026-09-15-deepgram-india-endpoint), but Soniox gates it behind monthly capacity commitments rather than self-serve access.

Also Speak

Details4 sources · 1 X post

Soniox launched a regional deployment in India so customers can run its STT and TTS APIs with processing and data kept in the country, with first access going to capacity-reservation customers.

India region

Soniox is now live in India with full data residency and lower latency. Customers can run Soniox STT and TTS entirely within India, with native-speaker accuracy for 10 major Indian languages and seamless mid-sentence language switching, including Hinglish. Initial access is limited due to constrained compute capacity. Learn more: https://soniox.com/blog/india-region

https://x.com/soniox_ai/status/2094331796158558713

Speaktext-to-speech6

NineNineSixModel providermodel release2026-09-05

NineNineSix ships Gepard 1.1, a GRPO pass that cuts unusable TTS outputs from 88 to 12

The jambonz listing filed the same week (2026-09-01-nineninesix-gepard-jambonz) names Gepard 1.0 rather than 1.1, and the gains come from NineNineSix's own benchmark and recogniser.

Details2 sources

Robustness update to the open Gepard text-to-speech model that drops and invents fewer words, with voice cloning and audio quality left essentially unchanged.

Gepard 1.1

LiveKitPlatformlaunch2026-09-05

LiveKit Agents 1.8.0 adds Vakyam TTS plugin

New TTS plugin.

Details1 source

livekit-agents 1.8.0 adds a Vakyam TTS plugin.

livekit-agents 1.8.0

DeepdubModel providermodel release2026-09-03net-new

Deepdub ships Phantom Z 3.4 Conversational TTS at 150 ms TTFA

Production-focused normalisation and 150 ms p95 TTFA at 48 kHz on 3 September — aimed at the account-number and Hebrew edge cases that break containment, not another demo voice.

Details1 source

Deepdub launched Phantom Z 3.4 Conversational, a multilingual real-time TTS model with 48 kHz audio, stronger English and Hebrew text normalisation, and a vendor-claimed p95 time-to-first-audio of 150 ms.

Phantom Z 3.4 Conversational

InworldModel providerlaunch2026-09-02

Inworld launches Realtime TTS-2 and TTS-2 Flash

Two-tier realtime TTS with prompt tags and LiveKit-native wiring. Flash TTFA (~25ms) is a Coval figure, not a bench we ran.

Details4 sources · 1 X post

Inworld launched a two-tier realtime TTS family with natural-language delivery steering, cloning, 200+ language cross-voice support, and on-demand pricing from $25 per million characters.

Realtime TTS-2

Introducing TTS-2, a new model from Inworld AI. TTS-2 is the most natural and expressive voice model we've ever made. It gives you precise control over how your voice agent speaks with natural language. Available now: https://t.co/UVCkO0j5BP

https://x.com/inworld/status/2051699438201282992
NineNineSixModel providerpartnership2026-09-01

jambonz adds NineNineSix's Gepard 1.0 as a built-in TTS vendor

Puts an open-weight TTS one setting away on self-hosted SIP stacks; the source is NineNineSix's own blog, and this sweep did not check a matching jambonz release note.

Also Connect

Details2 sources

Operators of the open-source jambonz telephony platform can pick Gepard 1.0 for phone calls at 8 kHz mu-law with one speech credential, billed at $5 per million characters.

Gepard TTS on jambonz

GradiumModel providermodel release2026-08-31

Gradium makes hard-case TTS the default after public beta

Closes the Jul beta on pronunciation of phones, emails, and IBANs before the later latency-focused Gradium drops in September.

Details2 sources

Gradium promoted a new default TTS model that passes 81.0% of a 500-sentence hard-case eval across five languages, with Coval TTFA P50 of 216 ms and no caller-side text normalisation.

Gradium TTS (hard-case default)

Showavatars, video1

TavusPlatformlaunch2026-09-01

Tavus gives PALs Browser Use for live product walkthroughs

The product shifts a conversational avatar from explaining a workflow to operating the browser in front of the user, joining voice interaction with computer-use behaviour.

Also OrchestrateThink

Details1 source · 1 X post

Tavus PALs can open a product, share their screen, click through the experience, and guide a person while speaking live.

Browser Use for PALs

Your PAL can now show, not just tell. Introducing Browser Use: PALs can open your product, share their screen, click through it, and guide someone through the experience while talking with them live. Product demos, onboarding, setup, training, and support now each come with a personal walkthrough.

https://x.com/tavus/status/2094826372787323180

ThinkLLMs, speech-to-speech2

PhonelyModel providermodel release2026-09-01

Phonely launches Alma, a voice LLM trained on more than 10M phone calls

A vertical phone-agent vendor ships its own think-layer model instead of only wrapping GPT-class APIs; latency and price claims are vendor-run against GPT-4.1 and GPT-5.4.

Also SpeakHear

Details3 sources

Phonely opened Alma, its own large language model for voice agents trained on more than 10 million real phone conversations, claiming sub-185 ms time to first token and a $0.55 blended cost per million tokens.

Alma

BolnaPlatformlaunch2026-08-31

Bolna adds a Realtime engine for OpenAI Realtime and Gemini Live agents

The S2S path excludes Bolna's graph and multilingual agents, the two features its India-first pitch leans on, so most Indian-language deployments stay on the cascaded stack for now.

Also Orchestrate

Details2 sources

Bolna agents can run on a single speech-to-speech model, OpenAI Realtime or Gemini Live, instead of separate STT, LLM, and TTS steps.

Bolna speech-to-speech agents · vendor OpenAI; Google

Orchestrateframeworks, evals7

LiveKitPlatformlaunch2026-09-05

LiveKit Agents 1.8.0 (Python + JS) moves traces to OTel GenAI conventions with in-process PII filtering

Breaking observability contract for Agent Insights and third-party GenAI backends across Python and Node.

Details2 sources

livekit-agents and @livekit/agents 1.8.0 align spans, metrics, and logs with OpenTelemetry GenAI semantic conventions and move PII filtering in-process; the Node core package ships the same 1.8 line.

livekit-agents / @livekit/agents 1.8.0

SoundHound AIPlatformacquisition2026-09-04

SoundHound AI completes its LivePerson acquisition and folds LivePerson's messaging platform into OASYS

The April agreement (2026-04-21-soundhound-liveperson, W17) valued LivePerson's equity at $43 million; after the 4 September close, SoundHound's human-assisted resolution (2026-09-16-soundhound-human-assisted-resolution, W38) and OASYS Edge (2026-09-24-soundhound-oasys-edge, W39) are the first product filings from the combined company.

Also Connect

Details2 sources · 1 X post

LivePerson is now an indirect wholly owned SoundHound subsidiary, and its digital messaging platform is being integrated into SoundHound's OASYS agent system across voice, web, mobile, SMS, and social channels.

LivePerson acquisition (completed)

SoundHound AI Completes Acquisition of LivePerson, Creating a World-Leading Omnichannel Conversational AI Powerhouse

https://x.com/SoundHound/status/2095880685332398472
WonderfulPlatformfunding2026-09-02

Wonderful raises $550M Series C at a $5B valuation

The round is a later financing step after Wonderful's $150M Series B in W11 and is tied to the company's shift from isolated AI use cases to an enterprise-wide operating system.

Also UseSpeak

Details2 sources · 1 X post

Wonderful raised $550 million in Series C funding at a $5 billion valuation to expand its AI operating system and global enterprise deployment.

Wonderful AI OS

Wonderful has raised a $550M Series C at a $5 billion valuation. This latest funding was led by @insightpartners, with @salesforce joining as new investors and @IndexVentures, @BessemerVP, @IVP, @VineVenturesLP, and @9yardscapital returning. In 20 months since launch, we're in production with hundreds of agents, systems, and workflows across a dozen verticals in over 30 markets. With this new capital, we’ll be overdelivering for our customers, hiring hundreds of cracked engineers, FDEs, and former founders, and expanding our product capabilities across the entire enterprise. Stay tuned 👀

https://x.com/wonderful_ai/status/2095125312929464388
LiveKitPlatformlaunch2026-09-02

LiveKit Cloud runs staging deployments of a voice agent beside production

Testing a change on real LiveKit Cloud infrastructure previously meant a second project with its own keys and secrets to keep in sync. Agent Simulations (2026-10-08-livekit-agent-simulations-ga, W41) later added scripted tests before any deployment; non-production deployments cover the live-call check that follows.

Details2 sources · 1 X post

Non-production deployments run the same container image in the same project with the same secrets under another name, so teams can try a model or prompt change on LiveKit Cloud before it reaches callers.

Agent deployments (non-production)

Agent deployments are live on LiveKit Cloud. Your voice agent can have a staging deployment running next to production. Same image, same project, same secrets, different name. Test a change on real infrastructure, then promote that exact build. Read more: https://t.co/3rmy9DnEAK

https://x.com/livekit/status/2095180824186425695
BolnaPlatformpartnership2026-09-02

Bolna lists its voice-agent platform on AWS Marketplace

Drawing down AWS commitments removes a procurement step for larger buyers; PolyAI took a similar route on 6 October with Dialog-RSN-1 on SageMaker (2026-10-06-polyai-dialog-rsn-1-aws-sagemaker), though PolyAI ships a self-hosted model and Bolna a hosted platform.

Details3 sources

Enterprises can buy Bolna Pilot and Enterprise contracts, monthly or annual, through AWS Marketplace and bill them against AWS committed spend.

Bolna on AWS Marketplace · vendor AWS

Retell AIPlatformlaunch2026-09-01

Retell ships native pre-call and post-call workflows

CRM glue at the call boundary without Zapier/n8n. Pre-call budget: 1 minute inbound/web, 5 minutes before an outbound dial.

Details1 source

Retell shipped native pre-call and post-call workflows that pull CRM/helpdesk context into voice agents and write outcomes back, without Zapier or n8n.

Retell Workflows

VapiPlatformlaunch2026-08-31

Vapi rebuilds outbound Campaigns on a durable workflow engine

Batch outbound with event-sourced state and a skip webhook, instead of a blob/counter. Lists up to 10,000 contacts.

Details1 source · 1 X post

Vapi rebuilt outbound Campaigns on a durable workflow engine for lists up to 10,000 contacts, with scheduling windows, dynamic variables, and pre-dial skip webhooks.

Vapi Campaigns

Vapi Campaigns is live, rebuilt from the ground up.

https://x.com/Vapi_AI/status/2094832678235209734

Connecttelephony, channels6

LiveKitPlatformlaunch2026-09-05

LiveKit Agents 1.8.0 adds Twilio connector warm transfer

Warm transfer on Twilio connector.

Details1 source

livekit-agents 1.8.0 adds Twilio connector warm transfer.

livekit-agents 1.8.0

GupshupPlatformlaunch2026-09-03net-new

Gupshup Voice AI Platform: self-serve PSTN and WhatsApp voice beside messaging

In several markets the voice agent is a messaging company’s PSTN add-on, not a CCaaS replacement.

Details1 source

Gupshup launched a self-serve console for building and deploying AI voice agents across PSTN and WhatsApp voice alongside WhatsApp, RCS, and SMS, with a choice of STT, TTS, and LLM.

Voice AI Platform

GenesysPlatformlaunch2026-09-02

Genesys Agentic Virtual Agent: APT-2, end-of-turn detection, ElevenLabs and Deepgram

The CCaaS SKU enterprises already bought is getting frontier TTS, STT, and an action model on a published timetable, with ElevenLabs still later this fiscal quarter.

Also Orchestrate

Details1 source · 1 X post

at Xperience, Genesys announced Agentic Virtual Agent upgrades: Scaled Cognition APT-2, end-of-turn detection, Contextual Tuning, and integrations with ElevenLabs and Deepgram.

Agentic Virtual Agent (APT-2 + voice) · vendor ElevenLabs / Deepgram / Scaled Cognition

Genesys is advancing agentic orchestration for CX: context + governance, a smarter Agentic Virtual Agent

https://x.com/Genesys/status/2095182564998709342
ZendeskPlatformlaunch2026-09-01net-new

Zendesk makes voice AI agents generally available on Zendesk voice lines

Zendesk's own Talk base gets native voice agents billed per confirmed resolution rather than per seat, nine days before the 10 September Real-Time Voice Translation launch (2026-09-10-zendesk-realtime-voice-translation); Zendesk Contact Center customers still wait for a future release.

Also Orchestrate

Details3 sources

Zendesk opened generative voice AI agents to all customers with Zendesk voice (Talk), handling calls end to end in 53 languages and handing off to humans with transcript, summary, and intent on the ticket.

Voice AI agents

QuiqPlatformlaunch2026-09-01net-new

Quiq Voice Assist: live-call guidance that can send the customer a message

This assist product can send the voucher during the call, which is a different job from a whisper bar that only suggests text to the agent.

Details1 source

Quiq extended its AI Assistants to live voice: adaptive next-best-action plus the ability to send customer-facing messages (links, vouchers) on the same thread while the human stays on the line.

Voice Assist

LiveKitPlatformlaunch2026-09-01

LiveKit Connectors GA: Twilio Voice and WhatsApp without a SIP trunk

Teams already on Twilio Studio, or on WhatsApp calling, can attach a LiveKit agent without standing up a SIP trunk.

Details1 source · 1 X post

LiveKit launched Connectors that bridge LiveKit rooms to Twilio Programmable Voice and WhatsApp Business Calling without provisioning a SIP trunk, generally available on LiveKit Cloud.

LiveKit Connectors

LiveKit Connectors are generally available today. They bring calls from @twilio Programmable Voice and @WhatsApp Business Calling into a LiveKit room as a regular participant, so your agent can answer there with no SIP trunk to provision and nothing new to host.

https://x.com/livekit/status/2094905369063755830

Useagents in market5

GooglePlatformlaunch2026-09-03

Google brings Gemini Live voice workflows to Gmail, Docs, and Keep

Voice moves into three everyday Workspace surfaces rather than remaining a standalone assistant destination; availability is described as rolling out, not as a universal same-day release.

Also Think

Details3 sources · 2 X posts

Google is rolling out voice capabilities in Gmail, Docs, and Keep so users can search inboxes, organise thoughts, and brainstorm conversationally with Gemini Live.

Gemini Live in Gmail, Docs, and Keep · vendor Google Gemini

We’re bringing new voice capabilities to @GoogleWorkspace to help you tackle daily tasks. These are now rolling out across @Gmail, @GoogleDocs, and Keep to help you search your inbox, organize your thoughts, or brainstorm new ideas conversationally. Here's how you can use them to get things done:…

https://x.com/Google/status/2095589177341403244

Docs Live lets you create and edit a document with your voice. From brainstorming to polishing, collaborate with Gemini in real time to bring your ideas to life.

https://x.com/Google/status/2056784832261210589
CommurePlatformlaunch2026-09-03

Commure Ambient AI becomes available across the UK and EU

This is a regulatory and geographic expansion of an already deployed clinical-documentation product, not an autonomous diagnostic system: clinicians remain responsible for review and sign-off.

Also Hear

Details2 sources · 1 X post

Commure made Ambient AI available to clinicians across the UK and EU as CE-marked Class I medical-device software registered with the UK MHRA.

Ambient AI

Commure Ambient AI is now available for clinicians across the UK and EU.

https://x.com/CommureOS/status/2095554139296235559
Fireflies.aiPlatformlaunch2026-09-02net-new

Fireflies Voice Agents: recruiting, sales, and support calls from a knowledge base

A meeting-notes vendor running the live call is a different buyer motion from CCaaS or voice-agent infra.

Details1 source · 1 X post

Fireflies launched Voice Agents that conduct recruiting, sales, and support calls from a team knowledge base rather than only taking notes, claiming 40,000+ conversations across 2,100+ organisations in 97 countries since rollout.

Voice Agents

Introducing Fireflies Voice Agents: AI that handles repetitive conversations like screening interviews, sales discovery, customer support, user research

https://x.com/firefliesai/status/2095179947455480113
Bland AIPlatformother2026-09-01

Bland passes FedRAMP 20x Class A and opens its voice agents to US federal agencies

Bland announced its 20x Class A a month before ElevenLabs announced the same tier (2026-10-01-elevenlabs-fedramp-20x-class-a), and its 18-day timeline rests on self-hosting every model; no agency deployment is named yet, so this is market access rather than a go-live.

Also Orchestrate

Details2 sources · 2 X posts

Federal agencies can now buy Bland's self-hosted voice-agent platform for calls such as appointment confirmations, benefit status checks, and eligibility verification.

Bland for Government (FedRAMP 20x Class A)

Bland is now the first FedRAMP 20x Class A certified voice AI company in the US

https://x.com/usebland/status/2094833694607605957

Bland's voice AI infrastructure is now available to federal agencies.

https://x.com/usebland/status/2095557502796701746
SalesforcePlatformlaunch2026-08-31

Salesforce launches Autonomous Scheduling with Agentforce Voice

This is a concrete transaction flow on top of Agentforce Voice: the agent moves from answering a call to completing a dispatch-sensitive booking against live availability.

Also ConnectOrchestrate

Details1 source

Autonomous Scheduling with Agentforce Voice lets callers book appointments around the clock while agents check availability and select a qualified technician and territory.

Autonomous Scheduling with Agentforce Voice