hotwirednews Get GrokBot

Week 47

16–22 Nov 20262026-W47 · 9 filings · 9 companies · 5 net-new

In progress. New filings arrive every 6 hours.

The Orchestrate layer leads so far with 4 of 9 filings, from Pikaia, ElevenLabs, Baseten and Artificial Analysis.

Hearspeech-to-text1

DeepgramModel providermodel release2026-11-16

Deepgram expands Nova-3 with 36 new languages for batch and streaming

The coverage expansion pushes Nova-3 towards broader multilingual batch and streaming deployments; it is additive language availability, distinct from the 12 November Nova-3 quality-only updates.

Details1 source

Deepgram added 36 monolingual languages to Nova-3, making them available for both batch and streaming speech-to-text workloads.

Nova-3 36 new monolingual languages

ThinkLLMs, speech-to-speech1

Liquid AIModel providermodel release2026-11-16

Liquid AI releases d1.2 and d1.2-ultra with audio input and structured outputs

Audio is a real input modality and spoken-question path rather than incidental metadata, extending Liquid's earlier d1 audio and LFM3 rows into a newer structured-decision API; the new audio-task scores are internal evaluations and OpenRouter/Vercel distribution is only described as coming soon.

Also HearOrchestrate

Details2 sources · 1 X post

Liquid AI released d1.2 and d1.2-ultra, omnimodal decision models that process text, images, and audio and add five structured-output tasks.

d1.2 and d1.2-ultra

Today, we release d1.2 and d1.2-ultra, the state-of-the-art omnimodal decision models. d1.2 and d1.2-ultra both can process text, images, and audio with new output capabilities. #1 on the @ArtificialAnlys Decision Index (75)...

https://x.com/liquidai/status/2122346077231923698

Orchestrateframeworks, evals4

PikaiaModel providermodel release2026-11-17net-new

Pikaia releases Smart Turn v3.2-Interview for bilingual voice-agent turn detection

Pikaia adds an interview-domain, bilingual endpointing checkpoint to the open turn-detection ecosystem without changing Pipecat's small-model interface; its reported gains need independent reproduction, and the companion Pikaia-Duplex Pro v6 system remains a private-weight research write-up.

Also Hear

Details4 sources

Pikaia published an 8.0M open-weight end-of-utterance model fine-tuned for Mandarin and English spoken interviews and designed as a drop-in Smart Turn v3 checkpoint.

Smart Turn v3.2-Interview

ElevenLabsPlatformacquisition2026-11-16

ElevenLabs agrees to acquire Kotoba for Japanese voice AI

The acquisition gives ElevenLabs local Japanese speech research and deployment expertise across TTS, STT, and speech-to-speech, while the announced close remains subject to customary approvals and the expanded product roadmap is not yet specified.

Also HearSpeakUse

Details2 sources · 1 X post

ElevenLabs agreed to acquire Kotoba Technologies, a Japanese voice AI lab whose Koto model supports Japanese text-to-speech, speech-to-text, and speech-to-speech.

Kotoba Technologies

. @Kotoba_tech, the leading voice AI lab in Japan, is joining ElevenLabs. Together, we'll bring Kotoba's Japanese voice research to ElevenAgents, so more businesses can serve customers naturally in Japanese. Welcome to ElevenLabs, @noriyuki_kojima, @jungokasai, and team.

https://x.com/ElevenLabs/status/2122420600408866844
BasetenPlatformlaunch2026-11-16net-new

Baseten launches a customizable Voice Agent Stack and Realtime Voice API

Baseten is offering control of the full speech cascade and orchestrator rather than only another hosted endpoint, with a migration path for OpenAI-Realtime-compatible agents; the waitlist and private-preview status leave production access and independent performance evidence open.

Also ConnectThinkSpeakHear

Details1 source

Baseten launched a private-preview stack for running customizable, low-latency voice agents and a reference Realtime Voice API on its inference infrastructure.

Baseten Voice Agent Stack and Baseten Realtime Voice

Artificial AnalysisPlatformlaunch2026-11-16

Artificial Analysis adds Time to First Audio to its TTS benchmark

The metric makes audible onset a separate TTS comparison from throughput, price, or arena preference, extending Artificial Analysis's WDER-based Diarization Index (2026-11-13-artificial-analysis-diarization-index) with another voice-system failure mode; the measurements remain vendor-run.

Also Speak

Details2 sources · 1 X post

Artificial Analysis added a Time to First Audio comparison for Text-to-Speech models, measuring how quickly each model begins producing audible output.

Artificial Analysis Time to First Audio benchmark

Announcing Artificial Analysis Time to First Audio, measuring how quickly Text to Speech models start speaking.

https://x.com/ArtificialAnlys/status/2122374288653373870

Useagents in market3

TopCallsPlatformlaunch2026-11-16net-new

TopCalls launches managed AI voice agents for outbound sales calls

A managed calling layer can serve sales teams without requiring them to assemble carrier, CRM, and follow-up plumbing; TopCalls' homepage claims 11.3 million calls handled and advertises pricing from $0.25 per minute, both vendor statements.

Also ConnectOrchestrate

Details4 sources

TopCalls runs inbound and outbound AI calls for sales teams and BPOs, including lead qualification, appointment setting, and follow-ups.

TopCalls AI voice agents

TonePlatformlaunch2026-11-16net-new

Tone launches goal-oriented voice agents and expressive AI narration

Tone's claimed differentiator is strategic control of tonality, volume, speed, and pacing rather than another general voice catalogue; TechCrunch says it could not independently verify the technology, leaving the reported 600-millisecond latency and production claims provisional.

Also ThinkSpeak

Details2 sources

Tone offers goal-oriented voice agents for sales calls and a self-service text-to-speech platform for creators and authors.

Tone AI voice agents and narration platform

Sempre HealthPlatformlaunch2026-11-16net-new

Sempre Health launches Atlas for medication-support calls across life-science and payer workflows

Atlas is a service-as-software operating model for healthcare voice work: Sempre owns the call operations, tuning, auditing, and human escalation instead of selling only an agent-building tool; the scale and security claims are company-reported.

Also ConnectOrchestrateHear

Details3 sources

Atlas is an AI voice service that handles patient, payer, provider-office, and pharmacy calls for life-science hubs, health plans, and PBMs.

Atlas