hotwirednews Get GrokBot

Week 19

4–10 May 20262026-W19 · 17 filings · 16 companies · 2 net-new

The Orchestrate layer led the week with 6 of 17 filings, from Twilio, Gladia, Infinitus, ElevenLabs and 2 more.

Hearspeech-to-text1

KrispPlatformlaunch2026-05-06

Krisp VIVA 2.0 adds turn, interruption, and signal models for voice agents

Predictive turn and interruption models join voice isolation in one pre-STT SDK on 6 May — aimed at the demo-to-production audio gap, not another TTS or CCaaS SKU.

Also Orchestrate

Details2 sources

Krisp launched VIVA 2.0, a server-side CPU SDK that sits before STT with Voice Isolation v3, Turn Prediction v3, Interruption Prediction v1, and signal detectors for synthetic speech, gender, and accent.

Krisp VIVA 2.0

Speaktext-to-speech3

SpeechifyModel providerlaunch2026-05-09

Speechify launches Simba 3.0 with streaming and voice controls

The model behind Speechify's reader app (60M+ users, per Baseten) became a sellable API with Baseten serving it from launch; the later Simba 3.2 and 3.5 rows build on this English-only base.

Details1 source · 1 X post

Simba 3.0 is a streaming-native English TTS model with per-voice speaking-rate and ADV emotion controls.

Simba 3.0

TelnyxPlatformAPI2026-05-06

Telnyx opens xAI Grok TTS and STT on standalone speech APIs

Speech-API surface for Grok on Telnyx's private network; distinct from the August Edge Compute runtime ship already filed.

Also Hear

Details1 source

Telnyx extended Grok beyond Voice AI Assistants to its TTS and STT APIs, with five expressive TTS voices and the xai/grok-stt transcription model.

xAI Grok TTS and STT API · vendor xAI

KyutaiModel providermodel release2026-05-04

Kyutai expands Pocket TTS to six languages without a GPU

An open 100M-parameter TTS release makes multilingual local inference available without requiring a GPU, extending Pocket TTS from an English model into a six-language speech engine.

Details2 sources · 1 X post

Pocket TTS's 100M-parameter models now generate speech in English, French, Spanish, Portuguese, Italian, and German, with the release open-sourced.

Pocket TTS multilingual models

Pocket TTS goes multilingual! Now you can use our 100M-parameter models to generate speech in six languages, fast enough to run real-time without a GPU. We also improved the quality of the English model while keeping the same size. And all of this is open-source.

https://x.com/kyutai_labs/status/2051317316713894368

ThinkLLMs, speech-to-speech1

OpenAIModel providermodel release2026-05-07

OpenAI ships GPT-Realtime-2 with Translate and Whisper companions

Splits realtime voice into reason / translate / transcribe primitives with GPT-5-class agent reasoning on the S2S path — token-priced Realtime-2 vs minute-priced companions changes how builders meter agents.

Also HearSpeak

Details2 sources

OpenAI released three Realtime API audio models: GPT-Realtime-2 (GPT-5-class speech-to-speech reasoning), GPT-Realtime-Translate (70+ input / 13 output languages), and GPT-Realtime-Whisper (streaming transcription), moving real-time audio beyond simple call-and-response.

GPT-Realtime-2, GPT-Realtime-Translate, GPT-Realtime-Whisper

Orchestrateframeworks, evals6

TwilioPlatformlaunch2026-05-07net-new

Twilio Forward introduces Ola as a communication and control layer for AI agents

Twilio is extending from communications infrastructure into the operator/control plane for personal and software agents; the post names Claude and OpenClaw but does not state a general-availability date or product endpoint.

Also ConnectUse

Details1 source · 1 X post

Twilio Forward introduced Ola, a single surface for managing agents across Claude, OpenClaw, and other runtimes, with controls for directing agents and approving sensitive actions.

Twilio Ola

Today Twilio Forward is introducing Ola — the communication and control layer for your AI agents. One surface to manage all of your agents, across Claude, OpenClaw, and more. Direct your agents, approve sensitive actions before they execute, and see everything they did on your behalf.

https://x.com/twilio/status/2052463368230019474
GladiaModel providerAPI2026-05-07

Gladia upgrades live and pre-recorded summarisation after Audio-to-LLM release

The update gives builders explicit output-shape controls for speech analysis while linking live and recorded transcription to Gladia's Audio-to-LLM path; the release post uses the spelling Audio-to-LMM, while Gladia's changelog and docs use Audio-to-LLM.

Also Hear

Details2 sources · 1 X post

Gladia added general, concise, and bullet-point summarisation styles to both pre-recorded and live transcription workflows.

Audio-to-LLM summarization

Big one for our API: following the release of Audio-to-LMM, Summarization has been upgraded for both pre-recorded and live transcription. 3 styles to pick from, defaulting to general when none is specified: general, concise, bullet_points. For setup and examples see the docs

https://x.com/gladia_io/status/2052357781211267223
InfinitusPlatformlaunch2026-05-06

Infinitus launches Lens to evaluate every healthcare conversation

Lens treats voice-agent quality, compliance, and human-agent monitoring as one healthcare conversation dataset rather than a sampled contact-centre review; the coverage and timing claims are from Infinitus.

Also HearUse

Details1 source · 1 X post

Lens evaluates every patient, provider, and payor interaction across voice, chat, SMS, and email, whether led by a human or an AI agent.

Lens

NEWS: Today we're launching Lens, the first solution to evaluate every healthcare interaction led by AI *or* humans: https://t.co/qeNduHcnpS

https://x.com/InfinitusAI/status/2052358567202640263
ElevenLabsPlatformlaunch2026-05-06

ElevenLabs adds images, files, audio notes, contacts, and locations to ElevenAgents

A voice-agent platform can now complete workflows that depend on documents, photos, or location data without sending the customer to a separate form or human queue.

Also Connect

Details1 source · 1 X post

ElevenAgents can process those modalities across WhatsApp and the web widget while preserving conversation context across channels.

ElevenAgents new modalities

Introducing new modalities for ElevenAgents Your customers don't just talk or type. They send photos, files, voice notes, and locations, and reach out across channels. Now your agents handle all of it.

https://x.com/ElevenLabs/status/2052022097283567746
SoundHound AIPlatformlaunch2026-05-05

SoundHound launches OASYS self-learning agent orchestration platform

Platform unveil on 5 May underpins the later HAR and Edge ships already on file — without it the SoundHound 2026 arc starts mid-year.

Also ConnectUse

Details2 sources

SoundHound AI launched OASYS (Orchestrated Agent System), a platform that builds, orchestrates, evaluates, and improves fleets of multilingual voice and agentic AI agents across phones, chat, kiosks, and in-vehicle channels.

OASYS (Orchestrated Agent System)

BolnaPlatformlaunch2026-05-04

Bolna launches Graph Agents for node-based voice call flows

Bolna's 1 October 2026 case post credits graph agents with a 30-minute wealth-advisory call on prompts of five lines or fewer per node; that zero-hallucination figure is the vendor's own.

Details2 sources

Bolna builders can now model a call as a graph of nodes, each with its own prompt and transition rules, instead of one long prompt.

Bolna Graph Agents

Connecttelephony, channels2

TwilioPlatformlaunch2026-05-06net-new

Twilio Agent Connect GA bridges AI runtimes to Voice and Messaging

Model-agnostic last-mile into Twilio Voice and SMS without per-provider WebSocket glue — complements LiveKit Connectors' later Twilio Voice bridge rather than duplicating it.

Also Orchestrate

Details2 sources · 1 X post

Twilio made Agent Connect generally available: a self-hosted, model-agnostic orchestration SDK that connects OpenAI, Azure, Bedrock, Anthropic, LangChain, or in-house agents to Twilio Voice and Messaging with streaming, barge-in, memory, and human handoff.

Twilio Agent Connect

🚨 ANNOUNCING CONVERSATION INTELLIGENCE 🚨 that uses generative AI Language Operators to analyze live voice and messaging conversations in real time. It also natively integrates with Conversation Orchestrator and Conversation Memory to enrich analysis with history and enterprise knowledge.

https://x.com/twilio/status/2052069119021973676
VobizPlatformfunding2026-05-05

Vobiz raises $1M seed led by Piper Serica

Dates the seed behind telephony already described as live the prior weeks — 1600-series numbers and 24 kHz plus in-pipeline redaction — rather than funding a model lab.

Details3 sources · 1 X post

Vobiz announced a $1 million seed round led by Piper Serica VC Fund to expand engineering and go-to-market for its AI-native telephony platform.

Vobiz

Today we raised $1M in seed funding led by @PiperSerica

https://x.com/Vobiz_Ai/status/2051561446862192810

Useagents in market4

CommurePlatformpartnership2026-05-07

Willis Knighton picks Commure as enterprise ambient scribe partner

Named MEDITECH Expanse ambient go-live with pilot adoption stats — Commure’s enterprise ambient path outside Epic-heavy accounts.

Also Hear

Details2 sources

Willis Knighton Health selected Commure Ambient AI as its enterprise ambient scribe after a seven-specialty pilot on MEDITECH Expanse, citing reduced charting time for 88% of participating providers.

Ambient AI × Willis Knighton Health

ParloaPlatformpartnership2026-05-05

Parloa becomes a certified solution on the Five9 CX Marketplace

Five9 relaunched its own Voice AI Agents on 23 June (2026-06-23-five9-voice-ai-agents), so Five9 customers now have a native product beside this marketplace listing; neither release compares the two.

Also Connect

Details1 source

Five9 customers can add Parloa AI agents to their Intelligent Cloud Contact Center, with existing routing, escalation, and human-agent operations carried into agent training.

Parloa on the Five9 CX Marketplace

Assort HealthPlatformlaunch2026-05-05net-new

Assort Health launches Activate for proactive patient outreach

Activate extends Assort's inbound voice-agent base into multi-step outbound care workflows rather than stopping at a reminder or a single answered call; the provider count is a company-reported figure.

Also ConnectOrchestrate

Details1 source · 1 X post

Assort Activate runs outbound patient campaigns across voice, SMS, and email until an appointment is booked, a referral is closed, or a balance is paid.

Assort Activate

Most healthcare AI is built to respond. Assort is built to act. Today, we're officially launching Assort Activate: a new standard for proactive patient outreach. Here's what makes it different: 🌐 True omni-channel engagement: Reach patients across voice, SMS, and email. Not just whichever channel is easiest. 🔄 End-to-end campaigns: Not just a reminder. Not a nudge. Activate runs the full outreach cycle until the appointment is booked, the referral is closed, or the balance is paid. 🧠 Built on Assort's industry-leading AI agents platform: Every Activate campaign inherits our specialty-trained models, Patient Journey Memory, and the precision to handle complex care scenarios out of the box so that actions are handled end-to-end during a single interaction. For the first time, AI doesn't just answer the phone. It stays with the patient. Read more in Fierce Healthcare: https://t.co/0yOJNSYy0n And stay tuned, as we'll be diving deeper into how Activate works and the real results our customers are already seeing.

https://x.com/assort_health/status/2051706083735916649
SierraPlatformfunding2026-05-04

Sierra raises $950M in Series E at a valuation above $15B

The round gives a voice-agent and customer-experience platform additional capital for its stated relationship-management strategy; the valuation and total-capital figures come from Sierra's announcement.

Also Orchestrate

Details1 source · 1 X post

Sierra said new and existing investors led by Tiger Global and GV are investing $950 million to expand its AI-agent customer-experience platform.

Series E

Series E - raising $950 million from new and existing investors, led by @Tiger_Global and @GVteam, at a valuation of over $15 billion. Sierra now has more than $1 billion to invest in becoming the global standard for companies transforming their customer experiences with AI. https://t.co/JjPvXhtopD

https://x.com/SierraPlatform/status/2051314764798943456