hotwirednews Get GrokBot

Week 9

23 Feb – 1 Mar 20262026-W09 · 14 filings · 13 companies · 8 net-new

The Orchestrate layer led the week with 4 of 14 filings, from Daily / Pipecat, IBM and Talkdesk.

Hearspeech-to-text3

GladiaModel providerlaunch2026-02-26net-new

Gladia makes PII redaction available for pre-recorded transcription

Redaction before persistence addresses a compliance boundary in transcript pipelines, where names, payment data, and identifiers can otherwise spread into downstream systems.

Details2 sources · 1 X post

Gladia made PII redaction available for pre-recorded transcription, replacing sensitive entities in transcripts before they are stored or returned.

PII redaction

PII redaction is now live in Gladia! This release is a key step towards building a secure, compliance-first voice AI engine for healthcare, finance, insurance, and other regulated environments. Available for pre-recorded transcription. For more: https://t.co/HclhPzEhdB https://t.co/3aBlsNMZHq

https://x.com/gladia_io/status/2027005735599849793
ConvoZenModel providermodel release2026-02-24net-new

ConvoZen ships Akshara STT and Ragini TTS for Indian telephony

India-telephony STT/TTS with a claimed 32% average WER cut on ConvoZen's own Indic bench — useful for BFSI/healthcare on-prem, but benches are vendor-run.

Also SpeakOrchestrate

Details2 sources

ConvoZen unveiled Akshara (speech-to-text) and Ragini (text-to-speech) at its Bengaluru AI Summit: telephony-first models trained on Indian call audio for Hinglish and regional languages, with claimed lower WER versus global baselines on the company's Indic bench.

Akshara + Ragini

ArgmaxPlatformlaunch2026-02-24

Argmax makes speaker-aware real-time transcription GA on iOS and macOS

Speaker labels move multi-party transcription into the on-device Apple SDK path instead of leaving diarisation as a separate post-processing step.

Details2 sources · 1 X post

Argmax made real-time transcription with speaker labels generally available on iOS and macOS through Argmax SDK 2.

Real-time Transcription with Speakers

Real-time Transcription with Speakers is now generally available on iOS and macOS! Details for installing or simply testing Argmax SDK 2 are in the comments. https://t.co/vYZjrwYzbO

https://x.com/argmax/status/2026392373887119411

Speaktext-to-speech2

WonderfulPlatformpartnership2026-02-26net-new

Wonderful adds Deepdub eTTS to enterprise multilingual voice agents

Enterprise agent platform wires emotive TTS from a localisation specialist — millions-of-calls claim is Wonderful's; Deepdub is the speak layer, not a new foundation model.

Also Use

Details1 source

Wonderful and Deepdub announced a partnership integrating Deepdub's Emotive Text-to-Speech into Wonderful's enterprise AI agent platform for emotionally nuanced, multilingual customer-facing voice at production scale.

Wonderful × Deepdub eTTS · vendor Deepdub

InworldModel providerlaunch2026-02-25

Inworld opens text-prompted Voice Design in research preview

Text-to-voice creation removes the need for a reference recording at the first design step, opening voice prototyping to users without source audio.

Details2 sources · 1 X post

Inworld's Voice Design feature generates up to three voice previews from a text description and makes selected voices available through its library and API.

Voice Design

You can now design a voice from a text description, no audio samples required. Describe the voice you have in mind and the model generates three voice previews to compare. Save the ones you like to your voice library and use them via API. Works in 15+ languages. Now available in our Portal as a research preview. Learn more: https://t.co/37cGl9HODE Try it out: https://t.co/YUVUEFMWFQ

https://x.com/inworld/status/2026703990625362341

Orchestrateframeworks, evals4

Daily / PipecatPlatformAPI2026-03-01

Pipecat Context Hub 0.0.5 splits compound queries into parallel sub-searches

A query that pairs, say, a TTS service with a transport can return hits for both halves; the notes give no relevance measurement for the interleaving.

Details1 source

Pipecat Context Hub 0.0.5 splits queries joined by " + " or " & " into sub-concepts, searches them in parallel, and interleaves the results.

Pipecat Context Hub 0.0.5

Daily / PipecatPlatformAPI2026-02-28

Pipecat Context Hub 0.0.4 adds get_hub_status and rescales retrieval confidence

The release notes say always-low confidence labels led Claude to bypass the MCP tools, so the rescaling matters more for tool uptake than for ranking order.

Details1 source

Pipecat Context Hub 0.0.4 adds a get_hub_status tool for index health and normalises retrieval scores to a 0 to 1 range.

Pipecat Context Hub 0.0.4

IBMPlatformpartnership2026-02-24net-new

IBM names Deepgram first voice partner for watsonx Orchestrate

Deepgram becomes IBM's first embedded voice partner on watsonx Orchestrate on 24 February — a month before the ElevenLabs TTS/STT add, so IBM's voice stack is multi-vendor from early on.

Also HearSpeak

Details3 sources

IBM and Deepgram announced a collaboration embedding Deepgram speech-to-text and text-to-speech into watsonx Orchestrate, making Deepgram IBM's first voice partner for building voice agents and voice-enabled workflows.

watsonx Orchestrate × Deepgram · vendor Deepgram

TalkdeskPlatformlaunch2026-02-23net-new

Talkdesk launches CXA Automation Flows for cross-system voice-agent workflows

The feature extends voice-agent work beyond a conversation into governed execution across CRM and other systems of record.

Also ConnectUse

Details2 sources · 1 X post

Talkdesk launched Automation Flows, a code-free orchestration layer for CXA that connects customer interactions and AI agents to multi-step backend workflows.

CXA Automation Flows

We're launching 2 updates to Talkdesk CXA to help you deploy the best AI-powered CX! 1️⃣ 𝐀𝐮𝐭𝐨𝐦𝐚𝐭𝐢𝐨𝐧 𝐅𝐥𝐨𝐰𝐬 drives autonomous orchestration across the backend 2️⃣ 𝐀𝐈 𝐀𝐠𝐞𝐧𝐭𝐬 𝐟𝐨𝐫 𝐄𝐦𝐚𝐢𝐥 now extends Autopilot to email Learn more: https://t.co/lHHeqSxgJh https://t.co/EXq2CJoCZz

https://x.com/Talkdesk/status/2025933866465100252

Connecttelephony, channels3

Voice ElementsPlatformlaunch2026-02-27net-new

Voice Elements adds Conversational AI Pipeline for realtime LLM telephony

Brings packet-level LLM audio streaming into an existing .NET SIP/WebRTC toolkit three months before the OpenAI Realtime-specific Voice Nexus launch.

Also Orchestrate

Details1 source

Voice Elements shipped AI Live Streaming / Conversational AI Pipeline so existing .NET telephony apps can stream caller audio to voice LLMs over SIP, WebRTC, or SMS without rewriting call flows.

Conversational AI Pipeline (AI Live Streaming)

TelnyxPlatformpartnership2026-02-25net-new

Telnyx announces a LiveKit plugin for production voice agents

The integration connects a carrier's infrastructure and inference services to a widely used agent runtime, reducing the glue code needed for production telephony.

Also Orchestrate

Details2 sources · 1 X post

Telnyx announced a LiveKit plugin that brings its voice and inference services into LiveKit agents with carrier-native performance and multiple speech-provider options.

Telnyx LiveKit Plugin · vendor LiveKit

🚀 Announcing the Telnyx LiveKit Plugin 💻 Add enterprise Voice AI to LiveKit agents with just a few lines of code. ✨ Enjoy multiple speech providers, ⚡️ carrier-native performance, 💰 and cost savings. 🎯 Perfect for production-scale Voice AI applications. 📖 Learn more: https://t.co/VOi2Zs5pAt #VoiceAI #LiveKit #Developers

https://x.com/telnyx/status/2026792826185388393
LiveKitPlatformlaunch2026-02-23

LiveKit Agents 1.4.3 adds sip_headers on WarmTransferTask

Telephony transfer knobs filed as connect.

Details1 source

livekit-agents 1.4.3 adds a sip_headers parameter to WarmTransferTask for SIP transfers.

livekit-agents 1.4.3

Useagents in market2

Deutsche TelekomPlatformlaunch2026-02-25

Deutsche Telekom launches CoMind voice and chat bots on ElevenLabs voices, live at Phoenix Pharma Hungary

A carrier packages ElevenLabs voices as its own European B2B voice-bot platform five days before the Magenta AI Call Assistant network unveil (2026-03-02-deutsche-telekom-magenta-ai-call-assistant); the 60%→98% service-level jump is customer-reported across a whole Webex consolidation, so it is not attributable to the voice bot alone.

Also Orchestrate

Details3 sources

Deutsche Telekom launched Telekom CoMind, a modular conversational AI platform of voice and chat bots grounded in company knowledge and using ElevenLabs voice technology; Phoenix Pharma Hungary runs it with Cisco Webex in its pharmacy-ordering contact centre.

Telekom CoMind · vendor ElevenLabs

Slang AIPlatformfunding2026-02-24net-new

Slang AI raises a $36M Series B led by USVP for restaurant voice agents

The round funded the guest-memory and proactive features Slang shipped as the rebuilt Superhost on 5 October (2026-10-05-slang-ai-superhost); the 25-million-call dataset is the company's stated moat and is not independently audited.

Details3 sources

Restaurant voice-agent company Slang AI raised $36 million ($28 million equity and $8 million debt) led by US Venture Partners, taking total funding to $68 million.

Superhost