hotwirednews Get GrokBot

Week 17

20–26 Apr 20262026-W17 · 19 filings · 16 companies · 5 net-new

The Use layer led the week with 6 of 19 filings, from Casey’s, Speechify, Cognizant, Tavus and 2 more.

Hearspeech-to-text2

MOSIModel providermodel release2026-04-24

MOSS-Audio unifies speech and sound understanding

The 71.08 score comes from MOSI’s reported evaluation, not an independent benchmark; the public X announcement followed the first weights by roughly two weeks.

Details2 sources · 1 X post

MOSI releases an open model family covering speech, emotion, speakers, sound events, music, temporal grounding, and audio reasoning.

MOSS-Audio

Meet MOSS-Audio. A unified open-source model for real-world audio understanding, built to handle speech, emotion, speakers, sound events, music, temporal grounding, and reasoning in one system. In the reported evaluation, our 4B model outperforms many 7B–9B open models, and MOSS-Audio-8B-Thinking reaches 71.08 average accuracy. Strong results. Real-world audio. GitHub: https://github.com/OpenMOSS/MOSS-Audio HuggingFace: https://huggingface.co/collections/OpenMOSS-Team/moss-audio MOSI.AI: https://mosi.cn OpenMOSS: https://www.open-moss.com

https://x.com/MosiAI_Official/status/2047518393582952793
LiveKitPlatformlaunch2026-04-22

LiveKit Agents 1.5.6 adds Inworld and Smallest Pulse STT across Agents 1.5.5 and 1.5.6

Same-week Agents 1.5.5–1.5.6 hear ships collapse to one card; diarization and new STT providers stay on one hear row.

Details3 sources

livekit-agents 1.5.6 adds Inworld STT and SmallestAI Pulse STT with realtime streaming and batch transcription; the same ISO-week hear cluster folds 1.5.5 inference STT diarization with speaker_id and Deepgram flux-general-multi.

livekit-agents 1.5.6 (cluster with 1.5.5)

Speaktext-to-speech3

SonioxModel providermodel release2026-04-23

Soniox launches Text-to-Speech API across 60+ languages

Moves Soniox from STT-only into a priced STT+TTS stack at $0.70/generated-hour — first TTS SKU; v2 with audio tags follows in August.

Details1 source

Soniox launched a Text-to-Speech API for high-fidelity speech in 60+ languages with streaming generation, alphanumeric precision, mixed-language handling, and regional deployment in the US, EU, and Japan, priced at $0.70 per hour of generated speech.

Soniox Text-to-Speech

MOSIModel providermodel release2026-04-22

MOSS-TTS-Nano brings stereo voice cloning to edge devices

The claimed sub-100 ms response on CPU makes the 100M checkpoint a smaller deployment option than the February MOSS-TTS Family (2026-02-11-mosi-moss-tts-family); MOSI provides the latency figure.

Details2 sources · 1 X post

The open roughly 100-million-parameter model generates speech in 20 languages and clones voices from three to five seconds of reference audio.

MOSS-TTS-Nano

Introducing MOSS-TTS-Nano A real-time, edge-native TTS model that brings together low latency, high-fidelity stereo audio, multilingual generation, and deployability in a single 100MB parameters! Fast on pure CPU. Native 48kHz stereo. 20 languages. Sub-100ms first audio latency. Zero-shot voice cloning from 3–5 seconds of reference audio. Open-sourced now. GitHub: https://github.com/OpenMOSS/MOSS-TTS-Nano Demo:https://openmoss.github.io/MOSS-TTS-Nano-Demo/ HuggingFace: https://huggingface.co/OpenMOSS-Team/MOSS-TTS-Nano-100M ModelScope: https://modelscope.cn/models/openmoss/MOSS-TTS-Nano

https://x.com/MosiAI_Official/status/2047009231749165331
LiveKitPlatformlaunch2026-04-22

LiveKit Agents 1.5.6 adds Qwen 3 and Minimax TTS across Agents 1.5.5 and 1.5.6

Same-week Agents 1.5.5–1.5.6 speak ships collapse to one card covering xAI, Qwen 3, and Minimax TTS.

Details3 sources

livekit-agents 1.5.6 adds Qwen 3 TTS on Simplismart and new Minimax TTS models; the same ISO-week speak cluster folds 1.5.5 xAI TTS on the inference path.

livekit-agents 1.5.6 (cluster with 1.5.5)

Showavatars, video2

AnamPlatformpartnership2026-04-23

Anam adds its live face layer to Stream Vision Agents

The integration separates Stream’s orchestration from Anam’s generated face layer; docs dated 17 April precede the public 23 April X announcement, and no deployment numbers are given.

Also Orchestrate

Details2 sources · 1 X post

Anam integrated its generated avatar layer with Stream’s open-source Vision Agents framework, which supplies orchestration, state management, and the realtime pipeline.

Anam × Vision Agents integration · vendor Stream / Vision Agents

Anam is now integrated with @visionagents_ai @getstream_io's open-source multimodal agent framework handles orchestration, state management, and the realtime pipeline. Drop Anam in as the face layer and the agent joins the call with a live generated face that reacts to what it sees and hears. What you can build: - a tutor reading student body language + eye contact - a sales coach catching the micro-signals a rep misses - an onboarding agent that sees confusion before it escalates Docs: https://t.co/GYym6MT9oe

https://x.com/Anam__ai/status/2047315814143914259
LiveKitPlatformlaunch2026-04-20

LiveKit Agents 1.5.5 adds AvatarSession base class

Shared avatar session abstraction.

Details1 source

livekit-agents 1.5.5 adds an AvatarSession base class and warns on sync mis-wire for remote avatars.

livekit-agents 1.5.5

ThinkLLMs, speech-to-speech1

xAIModel providermodel release2026-04-23net-new

xAI ships grok-voice-think-fast-1.0 flagship voice agent via API

Flagship S2S API with Starlink production metrics (20% sales conversion, 70% support containment, 28 tools) — τ-voice top spot and those rates are xAI-published.

Details1 source

xAI released grok-voice-think-fast-1.0, a full-duplex speech-to-speech voice agent API model aimed at multi-step support and sales workflows, with background reasoning, structured data capture, 25+ languages, and claimed top placement on the τ-voice Bench.

Grok Voice Think Fast 1.0

Orchestrateframeworks, evals4

SierraPlatformacquisition2026-04-23

Sierra acquires Paris-based AI operations company Fragment

The acquisition expands Sierra’s agent-development presence in France, but the 23 April announcement gives no transaction value or closing conditions.

Details1 source · 1 X post

Sierra acquired Paris-based Fragment, whose co-founders join Sierra’s agent-development efforts in France.

Fragment acquisition

We're excited to announce that Sierra has acquired Paris-based Fragment, which helps businesses scale their operations using AI. Fragment's cofounders are passionate about building better, more human customer experiences with AI, and bring valuable strength to our agent development efforts in France.

https://x.com/SierraPlatform/status/2047175791939891547
InfinitusPlatformlaunch2026-04-23

Infinitus launches Studio no-code builder for healthcare voice agents

Healthcare no-code voice-agent builder with ARC compliance switching and vendor-claimed 40%/90% accuracy/speed gains — figures are Infinitus-stated.

Also UseThink

Details2 sources

Infinitus launched Studio, a healthcare-specific no-code builder for designing, simulating, and deploying voice AI agents with Agent Response Control guardrails for payors and pharmaceutical programmes.

Infinitus Studio

SoundHound AIPlatformacquisition2026-04-21net-new

SoundHound agrees to acquire LivePerson for $43M equity value

Dated omnichannel M&A for a platform already in the corpus on OASYS — voice plus digital messaging under one roof, still pending H2 2026 close.

Also Connect

Details1 source

SoundHound AI and LivePerson announced a definitive agreement for SoundHound to acquire LivePerson, combining SoundHound's voice and agentic AI with LivePerson's digital messaging platform.

LivePerson acquisition

Daily / PipecatPlatformAPI2026-04-21

Pipecat Context Hub 0.0.17 adds reranker model choice and live reranker state in get_hub_status

The changes target operators diagnosing a degraded hub from logs; because serve now fails fast on an empty index, a fresh install has to run refresh before serve.

Details2 sources

Pipecat Context Hub 0.0.17 lets operators pick the cross-encoder model, reports live reranker state in get_hub_status, and makes serve exit with a remediation hint when the index is unusable.

Pipecat Context Hub 0.0.17

Connecttelephony, channels1

Synthflow AIPlatformpartnership2026-04-21net-new

Synthflow AI plugs its voice agents into 8x8 Contact Center in a strategic partnership

Synthflow's first CCaaS route in the corpus came six months before its Zoom Phone and Zoom Contact Center listing, which Batteries Plus now runs across 700+ stores (2026-10-20-batteries-plus-synthflow-zoom-phone, W43); in April, 8x8 resale and App Store distribution were still roadmap items.

Also Use

Details1 source

Synthflow's voice agents now run inside 8x8 Contact Center, where 8x8 customers can set up AI answering assistants for calls, chat, and digital channels without developers.

Synthflow AI agents for 8x8 Contact Center

Useagents in market6

Casey’sEnterprisepartnership2026-04-23net-new

Casey’s expands SoundHound AI ordering agents across more than 2,600 stores

A named convenience retailer is extending voice ordering to more than 2,600 locations after 21 million reported guest interactions; both figures come from SoundHound’s 23 April release.

Also Connect

Details2 sources · 1 X post

Casey’s renewed and expanded its partnership with SoundHound AI, extending AI-powered ordering agents across more than 2,600 stores.

SoundHound AI ordering agents across Casey’s stores · vendor SoundHound AI

NEWS: Casey’s Expands Partnership with SoundHound AI, Leveraging Cutting-Edge AI Agents Across More Than 2600 Stores 🍕 https://t.co/wrH7emd12d

https://x.com/SoundHound/status/2047301471771758659
SpeechifyPlatformlaunch2026-04-22

Speechify opens Simba Voice Agents early access for businesses

Consumer TTS brand opens a B2B voice-agent SKU with a 1,000,000-minute early-access pool — early access, not GA, and the 50M-user figure is vendor-claimed.

Also Think

Details1 source

Speechify opened an early access programme for Simba Voice Agents, a business voice-agent brand built on the company's TTS, STT, diarization, and speech-to-speech stack, offering selected partners 1,000,000 minutes plus Forward Deployed Engineer support.

Simba Voice Agents

CognizantEnablerlaunch2026-04-22net-new

Cognizant launches Agentic Retail CX on Google Gemini Enterprise

SI-packaged Gemini Enterprise contact-centre SKU for retail on 22 April — unnamed operators, so Cognizant stays the company rather than inventing a client.

Also ConnectOrchestrate

Details1 source

Cognizant launched Agentic Retail CX, a contact-centre solution on Google Cloud Gemini Enterprise for CX that targets retail self-service containment, omnichannel engagement, and agent-assisted workflows.

Agentic Retail CX · vendor Google Cloud

TavusPlatformlaunch2026-04-21

Tavus brings PALs to the iOS App Store

The 21 April App Store release gives Tavus PALs a native mobile distribution channel; the post provides no usage or retention figures.

Also Think

Details2 sources · 1 X post

Tavus released PALs on the iOS App Store, where the companions can talk, listen, and respond like the user.

PALs for iOS

The PALs are now available on the iOS App Store. They can talk, listen, and respond like you. Now, where you are. Download today for free 👇

https://x.com/tavus/status/2046646316990885981
Slang AIEnterprisepartnership2026-04-21

Slang AI moves restaurant voice agents to Deepgram

The post names a production switch by Slang AI and publishes customer-specific error and response claims; Deepgram’s comparison cites Pipecat’s 1,000-sample STT benchmark, not an independent Slang evaluation.

Also Connect

Details1 source · 1 X post

Slang AI says it migrated its restaurant voice agents to Deepgram, reporting 43% fewer errors and 72% faster responses than Google.

Restaurant voice agents on Deepgram · vendor Deepgram

Thousands of restaurants rely on Slang AI to keep seats full every night and run their business. Restaurants that don’t use AI-enabled phone systems miss on average 150 calls a month, which is roughly $28K in lost revenue per year, per location. When a caller doesn't feel heard, they usually hang up and call somewhere else. But Voice AI needs to do more than just transcribe. It needs to actually understand people, accurately and reliably, in restaurant-trained environments that are noisy and fast-paced. @slang_ai migrated to Deepgram to achieve 43% fewer errors and 72% faster responses than Google¹ -- a noticeable difference for both Slang AI and its customers. Slang AI, powered by Deepgram. "Slang AI answers every call for thousands of restaurants. That only works if the Voice AI behind it is fast and accurate. Deepgram gave us a measurable improvement, and our restaurants notice the difference.” — Urim Choi, VP Product & Technology, Slang AI #VoiceAI #DeepgramForRestaurants #RestaurantTech ¹ - Pipecat's open-source STT benchmark (1,000 samples, 2026)

https://x.com/DeepgramAI/status/2046739586957918353
PolyAIPlatformpartnership2026-04-21

PolyAI connects voice agents to Tripleseat for hospitality enquiries

The integration puts voice-agent event enquiries into a named hospitality system, so the value is reservation capture rather than a generic model integration; no deployment volume or launch pricing is published.

Also Connect

Details2 sources · 1 X post

PolyAI integrated with Tripleseat so its voice agents can capture hospitality enquiries, including event reservations, in the event-management platform.

PolyAI × Tripleseat integration · vendor Tripleseat

PolyAI and @Tripleseat are partnering to expand voice AI across the hospitality industry. Working deeply in the hospitality world, we've seen event reservations become critical to maximizing revenue. The Tripleseat integration enables PolyAI to capture every inquiry directly in the most widely used event management platform in the world. Together, we enable restaurants, hotels, venues, and event-driven businesses to deliver seamless, always-on guest experiences, from handling inbound calls to managing reservations and private event inquiries. We're constantly adding integrations to help customers turn more inquiries into bookings while removing work from their staff. Learn more about our partnership here: https://t.co/tEI0TnYFmS

https://x.com/polyaivoice/status/2046591313089323288