hotwirednews Get GrokBot

Week 14

30 Mar – 5 Apr 20262026-W14 · 14 filings · 12 companies · 7 net-new

The Use layer led the week with 5 of 14 filings, from OpenAI, LINE WORKS, Speechify, Invoca and 1 more.

Hearspeech-to-text3

MicrosoftModel providermodel release2026-04-02

Microsoft ships MAI-Transcribe-1 multilingual ASR on Foundry

Foundry-priced multilingual ASR at $0.36/hr with a claimed 2.5× Azure Fast batch speed — FLEURS wins are Microsoft's own comparisons.

Details2 sources

Microsoft released MAI-Transcribe-1 on Microsoft Foundry and the MAI Playground, a multilingual speech-to-text model covering the top 25 product languages, priced from $0.36 per audio hour with claimed 2.5× batch speed versus Azure Fast transcription.

MAI-Transcribe-1

CommurePlatformlaunch2026-04-02net-new

Commure launches Dictation for clinical voice at the cursor

Adds modular clinical dictation beside ambient — same pattern as Suki Dictation later in August.

Also Use

Details2 sources

Commure introduced Dictation, an AI clinical voice product that places speech-to-text at the cursor inside clinical applications to speed documentation alongside its ambient scribe line.

Dictation

WillowModel providermodel release2026-04-01net-new

Willow introduces Atlas 1 speech-to-text for real-time dictation

Atlas 1 is the predecessor named in Willow’s 7 July Frontier post, so the April model anchors the replacement path; the original X announcement publishes no WER or latency.

Details1 source · 1 X post

Willow introduced Atlas 1 as a frontier speech-to-text model for real-time dictation.

Atlas 1

Introducing Atlas 1. Willow's new frontier speech-to-text model. It outperforms ElevenLabs, Deepgram, OpenAI, and more by a wide margin. Built on the first scalable, human-powered transcription infrastructure ever built for real-time dictation.

https://x.com/WillowVoiceAI/status/2039393905616310659

Speaktext-to-speech3

Resemble AIModel providerAPI2026-04-02

Resemble AI launches Custom Vocabulary for reliable pronunciations

Project-scoped pronunciation control addresses a production failure mode that is especially costly in regulated or branded speech, without requiring a model retrain for each term.

Details2 sources · 1 X post

Resemble AI's Custom Vocabulary lets teams lock pronunciations for brand, drug, clinical, and other domain-specific terms across voices and API calls.

Custom Vocabulary

TTS has gotten remarkably good at everything except the words that matter. Brand names, drug names, clinical terms -- one mispronunciation in pharma, healthcare, or insurance isn't a UX issue, it's a compliance issue. Introducing Custom Vocabulary: upload a reference clip, lock in the correct pronunciation, and every voice and every API call gets it right from that point forward. https://t.co/FYnfwWBvuy

https://x.com/resembleai/status/2039743903088423041
MicrosoftModel providermodel release2026-04-02

Microsoft ships MAI-Voice-1 TTS with Foundry custom voice cloning

Foundry TTS with few-second custom voice cloning at $22/1M characters and a claimed 60× realtime generation factor — pairs with same-day Transcribe-1 for a Microsoft voice stack.

Details1 source

Microsoft released MAI-Voice-1 on Microsoft Foundry with secure custom-voice creation from a few seconds of audio, claiming 60 seconds of speech generated per second of compute and pricing from $22 per 1M characters.

MAI-Voice-1

GradiumModel providermodel release2026-04-01net-new

Gradium opens a private beta of Phonon, on-device TTS that runs on a phone CPU

Phonon swaps Gradium's per-request cloud pricing for a fixed licence per device and model type, aimed at freemium apps and games where most users never pay. Access was still a private beta in October, and every benchmark is Gradium's own.

Details4 sources · 1 X post

Phonon is a text-to-speech model of about 100M parameters that runs offline on a CPU. Gradium fine-tunes one model per partner for a single voice, language, and use case, and ships it as a licensed binary inside the app.

Phonon

API-based voice interaction works great, but scaling it to millions of free users is another story. Gradium Phonon: natural voices, multilingual, voice cloning, running locally on a smartphone CPU. No server, no latency, no per-call cost. Game devs, app builders: private beta is open, apply below ⬇️

https://x.com/GradiumAI/status/2039341282598813954

ThinkLLMs, speech-to-speech1

TavusPlatformpartnership2026-04-01

Tavus connects OpenPAL to OpenClaw for a 24/7 real-time assistant

An official 1 April integration ties a Tavus PAL to an external task agent; the post gives no availability, pricing, or implementation details.

Details1 source · 1 X post

Tavus connected OpenPAL to OpenClaw so its real-time AI assistant can send tasks to the claw-machine agent and return results.

OpenPAL × OpenClaw integration · vendor OpenClaw

Introducing the @openclaw x OpenPAL integration. Your 24/7 real-time AI assistant just got access to the most powerful claw machine on the internet. 🦞 Tasks go in, results come out, and your OpenPAL is the new version of you. More details in the thread:

https://x.com/tavus/status/2039406627112333726

Orchestrateframeworks, evals2

Daily / PipecatPlatformAPI2026-04-05

Pipecat Context Hub 0.0.14 adds check_deprecation and version pins on indexed chunks

Deprecation data here comes from heuristic parsing, which hub 0.2.1 replaces with Pipecat's machine-readable registry (2026-06-13-pipecat-context-hub-0-2-0).

Details1 source

Pipecat Context Hub 0.0.14 adds a check_deprecation MCP tool and records pipecat-ai version constraints on indexed examples and API chunks.

Pipecat Context Hub 0.0.14

Daily / PipecatPlatformAPI2026-03-31

Pipecat Context Hub 0.0.13 indexes TypeScript SDK symbols with tree-sitter

Coverage moves beyond Python to JavaScript and TypeScript clients, including PipecatClient; Swift, Kotlin, and C++ get language metadata and README discovery only, not symbol extraction.

Details2 sources

Pipecat Context Hub 0.0.13 parses TypeScript client SDKs with tree-sitter so search_api returns typed methods for client-side Pipecat code.

Pipecat Context Hub 0.0.13

Useagents in market5

OpenAIPlatformlaunch2026-04-02net-new

OpenAI brings ChatGPT voice mode to Apple CarPlay

The feature moves ChatGPT voice from the phone screen into supported cars, with rollout gated by iOS 26.4+ and CarPlay support.

Also Think

Details1 source · 1 X post

OpenAI began rolling out ChatGPT in Apple CarPlay to iPhone users on iOS 26.4 or later where CarPlay is supported.

ChatGPT voice mode in Apple CarPlay · vendor Apple CarPlay

ChatGPT is now available in CarPlay. The voice mode you know, now available on-the-go. Rolling out to iPhone users running iOS 26.4+ where CarPlay is supported.

https://x.com/OpenAI/status/2039748699350532097
LINE WORKSPlatformlaunch2026-04-02net-new

LINE WORKS begins selling NAVER CareCall for scheduled elder-care calls

The Izumo trial involved about 70 residents, while the service’s post-disaster safety checks and evacuation guidance remain future plans.

Also Orchestrate

Details3 sources · 1 X post

NAVER Cloud Japan’s CareCall phones older people on a schedule to check their safety and health and offer emotional support.

NAVER CareCall · vendor NAVER Cloud Japan

【プレスリリース🚩】 高齢者向けAI音声モニタリング電話サービス「NAVER CareCall」の販売を開始 ケアが必要な高齢者を対象に、AIが自動で定期的に電話をかけ、人と話すような自然な会話を通じて、安否確認や健康状態のモニタリング、心理的なサポートを提供します💡 詳細はこちら⏬

https://x.com/lineworks_jp/status/2039530353313140958
SpeechifyPlatformlaunch2026-03-31

Speechify launches Voice AI for Windows with text-to-speech and voice typing

One Windows surface combines text-to-speech with dictation across common work apps, while the Microsoft listing confirms a free install; the 1B+ reach is Speechify's claim and no adoption figure is published.

Also HearSpeak

Details3 sources · 1 X post

Speechify Voice AI for Windows reads text aloud and provides voice typing across desktop apps, with optional on-device inference.

Speechify Voice AI for Windows · vendor Microsoft Windows

Speechify Voice AI for Windows is here

https://x.com/SpeechifyAI/status/2039011378128670870
InvocaPlatformlaunch2026-03-31net-new

Invoca unveils AI Voice Agent that uses ad and web-journey data on inbound calls

Invoca's call-tracking record of which ad drove a call becomes context for the voice agent itself, an input most contact-centre agents lack; on 31 March the voice product was only announced, with beta promised for summer.

Also Connect

Details1 source

Invoca announced an AI Voice Agent that reads a caller's ad clicks and page visits before the conversation, qualifies the lead, routes high-intent callers to staff, and links each call outcome to its marketing source, alongside a wider SMS AI Messaging Agent.

AI Voice Agent

StellaPlatformlaunch2026-03-30net-new

Stella launches a voice-first anxiety companion app for iPhone

Mental-health voice companions carry a duty-of-care caveat that general assistants avoid, and Stella's site disclaims therapy and emergency use while charging $9.99 a month. With no named speech stack and only 18 ratings at capture, it is a small consumer signal, found six months after release through its Product Hunt launch.

Also Think

Details4 sources

People dealing with anxiety, stress, panic, or overthinking can press and talk to an iPhone companion that answers with conversational guidance and recalls what helped them before.

Stella: Voice Anxiety Support