hotwirednews Get GrokBot

Week 16

13–19 Apr 20262026-W16 · 23 filings · 18 companies · 8 net-new

The Use layer led the week with 7 of 23 filings, from Kotoba, Speechmatics, DeepL, Dairy Queen and 3 more.

Hearspeech-to-text2

SensoryModel providermodel release2026-04-15net-new

Sensory ships ultra-compact on-device STT for TFLM and NPUs

Edge STT at 2.7–13 MB with 37 languages targets offline/privacy devices — MAC and SRAM figures are vendor lab specs.

Details1 source

Sensory launched an on-device speech-to-text engine optimised for TensorFlow Lite Micro and NPUs including Arm Ethos-U55, with 37-language support and two footprints: a 2.7 MB domain-specific model and a 13 MB general-purpose model.

On-device STT engine

ArgmaxPlatformlaunch2026-04-15

Argmax opens WhisperKit and parts of its SDK as Argmax OSS

Argmax OSS turns an on-device Whisper deployment framework into a broader open SDK with speaker labelling and Qwen3-TTS playback; the 15 April release is partial rather than the whole Pro SDK.

Also OrchestrateSpeak

Details2 sources · 1 X post

Argmax released Argmax OSS with WhisperKit, SpeakerKit for speaker information, and TTSKit for realtime Qwen3-TTS generation and playback.

Argmax OSS

WhisperKit is now Argmax OSS! As part of our continued commitment to open-source, we are releasing part of Argmax Pro SDK, extending WhisperKit beyond speech-to-text. Argmax OSS now includes: - SpeakerKit: Add speaker info to your transcripts with the fastest implementation of Pyannote. - WhisperKit: One of the most popular frameworks to deploy Whisper with 6+ million monthly downloads. - TTSKit: Run Qwen3-TTS with real-time generation and playback for voice agents and content readers.

https://x.com/argmax/status/2044475694017552580

Speaktext-to-speech2

GoogleModel providermodel release2026-04-15

Google ships Gemini 3.1 Flash TTS with natural-language audio tags

Preview TTS with inline audio tags across 70+ languages and SynthID on every clip — Elo 1,211 is Artificial Analysis via Google's post, treat as third-party but vendor-framed.

Details2 sources

Google introduced Gemini 3.1 Flash TTS in preview on the Gemini API, Google AI Studio, and Vertex AI, adding natural-language audio tags for style, pace, and delivery across 70+ languages with SynthID watermarking on all audio.

Gemini 3.1 Flash TTS

Daily / PipecatPlatformlaunch2026-04-14

Pipecat 1.0.0 adds Mistral Voxtral TTS and NVIDIA cross-sentence stitching

Voxtral TTS and smoother NVIDIA Magpie turns expand open and hyperscaler speak options at GA.

Details3 sources

pipecat-ai 1.0.0 adds MistralTTSService for Voxtral TTS and enhances NvidiaTTSService with cross-sentence stitching within an LLM turn.

pipecat-ai 1.0.0

Showavatars, video3

LemonSliceModel providermodel release2026-04-17

LemonSlice 2.1 Flash targets sub-second avatar time-to-first-byte

Dated LiveKit Agents latency write-up for the pre-CWM stack. The later CWM-1 page also cites 471 ms, but that figure is interrupt frame time, not this Flash TTFT protocol.

Details1 source · 1 X post

LemonSlice 2.1 Flash is a video diffusion-transformer avatar model and inference stack that LemonSlice reports at 471 ms average time-to-first-byte, and 2.04 s average end-to-end when paired with third-party STT, LLM, and TTS.

LemonSlice 2.1 Flash

We did it! We built the fastest interactive avatar model Introducing LemonSlice-2.1 𝘍𝘭𝘢𝘴𝘩 ⚡ Here’s how we did it using @modal and @livekit 👇 (note: it was not easy)

https://x.com/LemonSliceAI/status/2045203651635843378
LiveKitPlatformlaunch2026-04-16

LiveKit Agents 1.5.3 adds Runway Characters avatar plugin

New avatar vendor for Agents. Class-2: folded quiet patch(es) 1.5.2 into this row.

Details2 sources

livekit-agents 1.5.3 adds a Runway Characters avatar plugin.

livekit-agents 1.5.3

Daily / PipecatPlatformAPI2026-04-14

Pipecat 1.0.0 adds LemonSlice avatar connect events

Presence events prevent talking to an avatar that has not joined yet.

Details3 sources

pipecat-ai 1.0.0 adds on_avatar_connected and on_avatar_disconnected on the LemonSlice transport when the avatar joins or leaves the Daily room.

pipecat-ai 1.0.0

ThinkLLMs, speech-to-speech2

LiveKitPlatformlaunch2026-04-16

LiveKit Agents 1.5.3 adds Cerebras LLM and xAI Grok inference

LLM provider cluster for 1.5.3. Class-2: folded quiet patch(es) 1.5.7, 1.5.14 into this row.

Details3 sources

livekit-agents 1.5.3 adds a Cerebras LLM plugin and xAI Grok LLM support on the inference path.

livekit-agents 1.5.3

Daily / PipecatPlatformlaunch2026-04-14

Pipecat 1.0.0 adds Inworld Realtime and WebSocket OpenAI Responses

Realtime Inworld and persistent Responses sockets become default think-layer paths at 1.0.

Details3 sources

pipecat-ai 1.0.0 adds InworldRealtimeLLMService (cascade STT/LLM/TTS with semantic VAD) and makes WebSocket OpenAIResponsesLLMService the Responses API default.

pipecat-ai 1.0.0

Orchestrateframeworks, evals6

LuaPlatformlaunch2026-04-16net-new

Lua ships Spaces supervisor agents to coordinate specialist fleets

Multi-agent supervisor with voice support on the same Agent OS that later added Browser/WhatsApp/Phone/Meetings transports — orchestration first, channel second.

Also Use

Details1 source

Lua launched Spaces: supervisor agents that reason over a fleet of specialised agents, delegate in one conversation, and synthesise a single reply across web, WhatsApp, Slack, Teams, and voice.

Spaces

LiveKitPlatformlaunch2026-04-16

LiveKit Agents 1.5.3 reuses realtime sessions and adds InstructionParts workflows

Realtime handoff reuse is the orchestration headline.

Details1 source

livekit-agents 1.5.3 reuses realtime sessions across agent handoffs when supported, and adds beta InstructionParts plus ToolSearchToolset and ToolProxyToolset.

livekit-agents 1.5.3

DaptaPlatformlaunch2026-04-16

Dapta Call Eval AI scores every voice-agent call against its conversation flow

Flow-aware post-call scoring ships inside an SMB agent builder rather than from a separate QA vendor such as Cekura or Coval (analytical comparison); Dapta publishes no accuracy figure for the evaluator, so its Accuracy Score is itself unvalidated.

Details1 source

Call Eval AI audits each call a Dapta voice agent handles against the builder's Conversation Flow and returns a weighted Accuracy Score plus a JSON list of failed blocks, missed intents, and CRM sync errors.

Dapta Call Eval AI

CloudflarePlatformlaunch2026-04-15net-new

Cloudflare adds experimental @cloudflare/voice to Agents SDK

Puts voice on the same Durable Object agent as text and tools without a separate framework — experimental, with Workers AI defaults and a Twilio adapter caveat on mulaw audio.

Also HearSpeakConnect

Details2 sources

Cloudflare released an experimental @cloudflare/voice package for the Agents SDK so the same Durable Object agent can take real-time voice over its existing WebSocket, with built-in Workers AI Flux/Nova 3 STT and Aura TTS plus a Twilio phone adapter.

@cloudflare/voice

Daily / PipecatPlatformlaunch2026-04-14

Pipecat hits 1.0.0 with Settings-pattern services and major removals

1.0 locks the builder-facing service and frame contract the rest of 2026 builds on.

Details3 sources

pipecat-ai 1.0.0 stabilises the Settings-pattern service APIs, async function calls with cancel_on_interruption=False, LLMMessagesTransformFrame, and removes deprecated pre-1.0 surfaces.

pipecat-ai 1.0.0

Bland AIPlatformlaunch2026-04-13

Bland adds agent-to-agent testing with simulated AI callers and success criteria

Simulated-caller testing, which Cekura and Coval sell as separate products, now sits inside Bland's own builder; Bland added A/B experiments and rebuilt evals in the October dashboard (2026-10-13-bland-dashboard-v2), and the April changelog gives no pricing or run limits.

Details3 sources

Teams define caller personas and success criteria, run the scenarios one at a time or in batches against their Bland voice agents, and review the transcripts.

Agent-to-Agent Testing

Connecttelephony, channels1

NTT Docomo BusinessPlatformlaunch2026-04-17net-new

NTT Docomo Business launches ANCAR Routing, an AI-IVR that routes spoken requests on Japanese toll-free numbers

At ¥10 per caller utterance, the release prices speech routing per request, which most IVR-replacement filings leave out. Two months later the same company turned to ElevenLabs for agents that answer calls by voice (2026-06-29-ntt-docomo-business-elevenlabs-contact-center, W27); same-week Kotoba on SoftBank (2026-04-17-kotoba-softbank-daredemo-ai) is the other Japanese carrier voice ship (analytical).

Also Hear

Details1 source

Callers to Free Dial and Navi Dial numbers say what they need instead of working through keypad menus, and the AI sends them to the right desk; a companion ANCAR Chat bot can take the caller over by SMS link.

docomo business ANCAR Routing

Useagents in market7

KotobaModel providerpartnership2026-04-17

Kotoba joins SoftBank Daredemo AI and ships Japanese–Korean S2S interpretation

Consumer/carrier distribution for Kotoba plus a named East-Asia S2S language pair beyond the May API alpha and August Agentforce JP filings.

Also ThinkSpeak

Details2 sources

Kotoba integrated its realtime interpretation app into SoftBank's Daredemo AI on 17 April 2026 and officially released a Koto-powered Japanese–Korean speech-to-speech model.

SoftBank Daredemo AI + Japanese–Korean S2S

SpeechmaticsPlatformpartnership2026-04-16

Speechmatics and thymia launch a health-signal platform from 15 seconds of speech

The launch points speech AI at clinical screening and monitoring rather than transcription alone; the 30-plus signals and 15-second window are vendor claims, and the post gives no clinical validation.

Also Hear

Details1 source · 1 X post

Speechmatics and thymia launched a platform that combines Speechmatics speech-to-text with thymia voice biomarkers to surface more than 30 health signals from 15 seconds of natural speech in real time.

Speechmatics × thymia health-signal platform · vendor thymia

BREAKING 🚨15 seconds of speech can now surface 30+ health signals. 🎙️🩺 Today, @Speechmatics and thymia are launching a joint platform that pairs medical-grade speech-to-text with clinical-grade voice biomarker intelligence, in real time, from natural speech.

https://x.com/Speechmatics/status/2044789883277869067
DeepLPlatformlaunch2026-04-16

DeepL launches Voice-to-Voice real-time spoken translation suite

Speech-to-speech translation suite with staggered GA/early-access SKUs and 40+ languages — Slator 96% preference is DeepL-commissioned.

Also ThinkConnect

Details3 sources · 1 X post

DeepL launched DeepL Voice-to-Voice for live spoken translation across meetings, mobile/web conversations, group QR sessions, and a Voice-to-Voice API, expanding DeepL Voice to over forty languages including all 24 official EU languages.

DeepL Voice-to-Voice

Today we launched Group Conversations for DeepL Voice. It’s a huge breakthrough in how groups of people can communicate across languages, in person. We’re very excited about it – and this demo from our Spring Launch shows why. Group Conversations means anyone with a mobile device, headphones and a QR code can follow, engage and participate in discussions, in their preferred language. That’s massive for frontline workers and their managers, for people attending conferences and speaking on panels, for in-person board meetings, and more. Group Conversations is available now for all DeepL Voice for Conversations and Voice for Meetings Business customers. Reach out to our team here: https://t.co/8Xu5iXAhMt #DeepLVoice #VoiceToVoice #VoiceForConversations #GroupConversations

https://x.com/DeepLcom/status/2049870219787583758
Dairy QueenEnterprisepartnership2026-04-16net-new

Dairy Queen expands Presto Voice AI drive-thru pilots to select franchisees

Named national QSR operator go-live path — corporate test into select franchisee expansion — for Presto Voice rather than an unnamed pilot.

Details1 source

American Dairy Queen Corporation partnered with Presto on drive-thru Voice AI after corporate-store tests, expanding the pilot to selected franchisees across the US.

Drive-thru Voice AI (Presto) · vendor Presto

Vodafone SpainPlatformlaunch2026-04-15net-new

Vodafone Spain launches IAgente voice and chat agents for SMEs and enterprises

A European carrier was selling a partner-built voice and chat agent to SMEs from April, five and a half months before BT Business's RingCentral-branded AI Receptionist (2026-10-01-bt-ai-receptionist, W40); the 100,000-SME rollout figure from October is Vodafone's own.

Also Connect

Details6 sources

Vodafone Empresas launched IAgente with partner Jusan, an omnichannel agent that answers phone calls, WhatsApp, and web chat for Spanish businesses, books appointments, and hands off to a person.

IAgente (Vodafone Empresas) · vendor Jusan

PhonelyPlatformfunding2026-04-15net-new

Phonely raises a $16M Series A led by Base10 for AI phone agents

Customer co-investors in a generalist voice-agent Series A are uncommon; Alma, Phonely's own voice LLM, follows on 1 September (2026-09-01-phonely-alma).

Details2 sources · 1 X post

Phonely raised $16 million led by Base10 Partners with Y Combinator and enterprise customers Etech Global Services, TSA Group, and Engage CX, bringing total funding to $19 million at a stated valuation over $100 million.

Phonely AI voice agents

We raised a $16M Series A to fix the biggest blind spot in business: phone calls. $2 billion is transacted over the phone every day. And most companies have zero insight into what's happening on those calls. Why? There's no analytics. No optimization. No way to know what's working and what's losing you money. So we built the optimization engine from scratch. Voice agents that sound human. A/B testing for phone calls. The most advanced voice AI analytics on the market. One customer generated $10M in net new revenue in three months. We've gone from hundreds of calls to millions. Outperforming the best human reps. Getting better every day. Every business leader keeps getting asked: how will you use AI? Now you have an answer. Start with the phone. Optimize your calls with @phonely_ai . -- Excited to partner with @Base10Partners, @ycombinator, and more to build the future of voice.

https://x.com/phonely_ai/status/2044564997779689896
TeslaEnterpriselaunch2026-04-13net-new

Tesla adds a “Hey Grok” wake word for hands-free Grok in its cars

A spoken wake word takes Grok off the touchscreen and makes it a hands-free in-car assistant, the base that the Summer Update (2026-07-21-tesla-grok-vehicle-controls-calls) later extends to calls and car controls; Tesla publishes no usage figures.

Also Think

Details4 sources · 3 X posts

Tesla's Spring Update 2026 (software 2026.14) lets drivers start a Grok conversation by saying “Hey Grok”, ask questions, and set location-based reminders, then end it by saying “Goodbye”.

Grok in Tesla: "Hey Grok" wake word (Spring Update 2026.14) · vendor xAI

https://x.com/Tesla/status/2043782316850573312

"Hey Grok” lets you ask a question or set a location-based reminder Enable via Controls > Audio

https://x.com/Tesla/status/2044472634818367950

Say “Hey Grok” to initiate a conversation with Grok in your Tesla To dismiss Grok, simply say “Goodbye” at the end

https://x.com/Tesla/status/2057133720994529601