Hearspeech-to-text8
LiveKit Agents adds AssemblyAI STT 3.7 Realtime Agent and Sarvam v4
The release expands named realtime transcription choices across both supported Agents packages, while the changelog publishes no comparative accuracy, latency, quota, or pricing results.
Details2 sources
LiveKit Agents 1.8.13 and the same-week JavaScript 1.10.1 release add AssemblyAI STT 3.7 Realtime Agent and Sarvam saaras:v4 paths for realtime speech input.
livekit-agents 1.8.13 / @livekit/agents 1.10.1
InworldModel providermodel release2026-11-18
Inworld open-sources Voice Isolator and SAVA with a TTS-2 report
Open weights make two production-oriented voice components available outside Inworld’s hosted stack: the realtime isolator uses 20 ms of look-ahead, while SAVA targets turn timing on one CPU thread; the TTS-2 report adds a reproducible research reference.
Also OrchestrateSpeak
Details6 sources · 1 X post
Inworld released Apache 2.0 code and weights for 48 kHz Voice Isolator and SAVA, a streaming model for speech and turn detection, alongside a technical report on Realtime TTS-2.
Voice Isolator, SAVA, and Realtime TTS-2 report
AssemblyAI ships STT-3.7 for realtime, agent, and recorded speech
A single model family now covers live speaker-labelled calls, agent-focused transcription, and batch audio, with published hourly prices and a LiveKit Inference distribution path; the accuracy and latency comparisons remain AssemblyAI’s own measurements.
Details2 sources · 2 X posts
AssemblyAI released the STT-3.7 family for live conversations, voice agents, and recorded audio, with speaker labels, background-speech handling, and 32-language streaming.
AssemblyAI-STT-3.7
Hearless updates voice-note transcription and dictation on mobile
The update joins passive voice-note transcription with active keyboard dictation in one consumer workflow, while the release notes provide no independent accuracy or model information.
Also Use
Details5 sources
Hearless updated its WhatsApp voice-note transcription app with a keyboard, longer-dictation preview, and direct transcription access for shared voice notes.
Hearless voice notes to text 1.0.15
ElevenLabs brings medical transcription to live voice-agent calls
Medical terminology and turn-aware handling target failure modes that can make live healthcare agents unsafe or force callbacks; the evaluation numbers are ElevenLabs internal tests.
Also OrchestrateUse
Details2 sources · 1 X post
Scribe v3 Live Medical adds clinical-language transcription to live patient, member, and provider calls in ElevenAgents.
Scribe v3 Live Medical
ElevenLabs lowers Scribe batch and realtime transcription prices
Lower list prices reduce the cost of testing and operating Scribe-based transcription and voice agents for non-Enterprise customers, while the unchanged Enterprise terms and absence of post-change volume data limit the market signal.
Details2 sources
ElevenLabs set Scribe Speech to Text at $0.20 per hour for batch audio and $0.22 per hour for realtime audio on self-serve and Business plans.
Scribe Speech to Text pricing
ElevenLabs previews an audio watermark provenance API
A machine-readable watermark check gives applications a way to screen audio provenance, but the endpoint is beta and its coverage is limited to watermarks the service can identify.
Also Orchestrate
Details2 sources
A beta endpoint checks uploaded audio for an ElevenLabs watermark and can return identifying generation metadata.
Content Provenance API
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
Orchestrateframeworks, evals11
Twilio adds a silence threshold control to Conversation Relay turn detection
A configurable silence threshold gives phone-agent builders a deployment-specific control alongside Twilio's semantics-aware Smart Turn path, but tuning remains workload-dependent and the release supplies no recommended values.
Also ConnectHear
Details2 sources
Conversation Relay now exposes `turnMinSilence`, a TwiML attribute that controls how long it waits before reporting that a caller has finished speaking.
Conversation Relay turnMinSilence
ParloaPlatformlaunch2026-11-18
Parloa adds custom metrics and threshold alerts to Lens
Business-defined outcomes turn Lens from a fixed evaluation catalogue into a monitoring layer for an organisation's own service goals; the 18 November article gives no access or pricing terms.
Also Use
Details2 sources · 1 X post
Parloa Lens now lets teams define business-specific agent metrics, test them against conversations, and receive alerts when thresholds are crossed.
Parloa Lens custom metrics and threshold alerts
Did an AI agent suggest another flight to a customer or complete the rebooking? Custom metrics in Parloa Lens measure outcomes against your own definition of success. Alerting helps spot when agents fall short. The team gets an email when the chosen thresholds are crossed. Alerts link directly to the relevant conversations... Read more: https://okt.to/NYVqm8
https://x.com/parloa_ai/status/2123364173950099963
AgoraPlatformlaunch2026-11-18
Agora launches an AI Developer Toolkit for voice-agent builders
Putting project setup, SDK selection, and voice-agent examples into coding-agent workflows shortens the path from prompt to a working Agora prototype; the stated 5–10 minute target is a launch aspiration, not an independent build-time measure.
Also ConnectThink
Details1 source · 1 X post
Agora’s AI Developer Toolkit combines coding-agent skills, a project CLI, recipes, and Agora SDKs to help developers build voice agents and other realtime applications.
Agora AI Developer Toolkit
TwilioPlatformlaunch2026-11-17
Twilio adds supervision operators for AI-agent conversations
Post-call task and hallucination checks move voice-agent quality review from anecdotal call listening toward measurable failure categories, although the feature is still public beta.
Also UseConnect
Details2 sources · 1 X post
Twilio added Conversation Intelligence operators that evaluate whether an AI agent completed a task and flag hallucinations after a conversation ends.
Conversation Intelligence AI agent supervision
Pinecall launches an open-source harness for voice AI agents
Pinecall combines a self-hostable runtime with a supervisory console and multiple call surfaces, although this dated release is primarily reliability and deployment maintenance rather than a new speech model.
Also ConnectUse
Details4 sources
Pinecall offers an Apache-2.0 voice-agent platform with a runtime, console, SDKs, and CLI for phone, web, and WhatsApp agents.
Pinecall open-source voice AI harness v0.11.43
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
Genesys makes agentic voice agents deployable beside existing contact centers
Keeping the existing carrier and contact-center platform lowers the migration barrier for voice-agent adoption; claimed customer outcomes come from Genesys deployments and are not independent evaluations.
Also ConnectUseThink
Details2 sources · 1 X post
Genesys introduced Standalone AI Agents, allowing organizations to deploy agentic virtual agents with existing contact centers, phone numbers, and carriers.
Genesys Standalone AI Agents and Genesys Arena
Two new products: Standalone AI Agents can be deployed on your existing contact center platform; Genesys Arena brings Agent Kits, voice personas and workflows to help teams prototype agentic experiences.
https://x.com/Genesys/status/2122692498103419100
ElevenLabs adds code-built tools and hosted MCP agent publishing
Code-defined tools and a staged MCP control plane make the ElevenAgents configuration surface more reproducible and automatable, but the changelog does not publish adoption, latency, or reliability results.
Also ConnectUse
Details3 sources
ElevenAgents builders can define tools in the Python or JavaScript SDKs and use a hosted MCP server to edit, test, and publish agents.
ElevenAgents tools as code and Hosted MCP
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 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
Connecttelephony, channels6
Truecaller and Bolna add identity and follow-up controls to AI voice calls
Caller identity and a post-call choice address trust before and after an automated conversation, while the 19 November release provides no independent evidence on answer rates or spam outcomes.
Also UseOrchestrate
Details2 sources
The partnership lets Bolna's AI callers show a business identity, reason for contact, optional AI watermark, and post-call actions through Truecaller for Business.
Truecaller for Business and Bolna AI-call transparency · vendor Bolna
Twilio adds STT and TTS provider failover to Conversation Relay
Provider-level failover turns a single-vendor speech outage into a recoverable call-path event, although builders still need to verify provider compatibility and the operational behaviour of the switch.
Also HearSpeak
Details2 sources
Conversation Relay can switch a live call to backup speech-to-text and text-to-speech providers when the configured provider fails.
Conversation Relay STT/TTS provider failover
Retell launches Dialer for AI-agent outbound campaigns
An AI-native dialer makes campaign throughput depend on call rules and agent capacity rather than human-representative headcount; Retell’s dated guidance still leaves deployment outcomes and pricing open.
Also OrchestrateUse
Details2 sources · 1 X post
Retell Dialer runs outbound campaigns with AI agents, automated retries and callbacks, CRM-connected outcomes, and voice and SMS follow-ups.
Retell Dialer
NiCEPlatformpartnership2026-11-18net-new
NiCE connects Cognigy voice agents with Microsoft Teams and Dynamics 365
The integration places a voice agent inside existing enterprise telephony and back-office workflows while preserving human collaboration, but the announcement gives no customer rollout, call-volume, or quality results.
Also OrchestrateUse
Details2 sources
NiCE Cognigy can run as an Intelligent Voice Agent in Teams Phone and collaborate with employees through Teams, Azure, and Dynamics 365 Customer Service.
NiCE Cognigy and Microsoft Teams Phone integration · vendor Microsoft Teams Phone, Azure, Dynamics 365 Customer Service
GoToPlatformlaunch2026-11-17net-new
GoTo extends AI Receptionist from phone calls to digital channels
Multichannel continuity makes the receptionist useful when a customer starts by phone and continues digitally, though the release provides no independent quality or deployment evidence.
Also UseOrchestrate
Details2 sources
GoTo expanded GoTo Connect AI Receptionist beyond phone handling to manage customer conversations across voice and digital channels.
GoTo Connect AI Receptionist
RingZeePlatformlaunch2026-11-16net-new
RingZee gives ChatGPT, Claude, and Grok a real phone number
A connector that gives personal agents a carrier-facing number makes callbacks and inbound messages part of the same assistant workflow as outbound errands; RingZee publishes no independent call-quality or model evaluation.
Also Use
Details4 sources
RingZee lets ChatGPT, Claude, and Grok place and receive calls through a user-owned number, with pay-as-you-go pricing.
RingZee phone connector
Useagents in market14
StrettoEnterpriselaunch2026-11-18net-new
Stretto puts Tara bankruptcy-intake voice agent into production
The story is a concrete production customer deployment for a native speech-to-speech model in a sensitive intake workflow, while AWS and Stretto do not publish independent accuracy, conversion, or safety metrics.
Also ConnectThinkSpeak
Details1 source
Stretto replaced web-form lead capture with Tara, a phone-based AI voice agent that conducts bankruptcy-intake interviews and qualifies leads.
Tara · vendor Amazon Nova 2 Sonic · delivered by CI&T
NapsterPlatformlaunch2026-11-18net-new
Napster launches Pulse for scalable AI video interviews
Pulse applies a voice-and-video agent to research at scale: the launch includes 97 interview templates and Napster’s own weekly interviews with more than 60 people, while the supplied evidence does not independently measure interview quality.
Also ThinkShowOrchestrate
Details3 sources
Napster Pulse has AI agents hold one-to-one video conversations with employees, customers, or partners and turn the transcripts into themes and findings for leaders.
Napster Pulse
InfoTalk launches CantoTalk, a free Cantonese AI hotline
A phone-only interface brings a screen-free voice assistant to any Hong Kong number through 2668-7000, with Cantonese-English code-switching and no sign-up; the service remains limited to everyday information and warns that its AI answers may be inaccurate.
Also ConnectThinkHearSpeak
Details3 sources
CantoTalk lets anyone with a Hong Kong landline or mobile number call 2668-7000 for spoken answers in everyday Cantonese, including Cantonese-English code-switching.
CantoTalk
Granola brings Voice Typing to Android
Android availability extends Granola’s cross-app dictation surface beyond desktop and iPhone on 18 November, while the feature remains an end-user transcription tool rather than a new speech model.
Also Hear
Details2 sources · 1 X post
Granola expanded its free Voice Typing feature to Android, letting users dictate in any app with no stated word limit.
Voice Typing for Android
ElevenLabs introduces ElevenAgents for Healthcare
This is a distinct agent product and customer-use-case announcement rather than a second medical-transcription model row; Scribe v3 Live Medical remains the underlying hear layer filed on 17 November.
Also HearOrchestrateConnect
Details2 sources · 1 X post
ElevenAgents for Healthcare lets teams deploy conversational agents for patient and clinician calls that can book appointments and check claims.
ElevenAgents for Healthcare
Checkers & Rally’s extends Hi Auto voice ordering through 2030
This is an operating deployment and expansion rather than a generic restaurant-AI claim: the release gives a live store count, bilingual ordering, POS action, human fallback, and customer-reported outcome measures.
Also ConnectSpeak
Details1 source
Checkers & Rally’s extended its Hi Auto partnership through 2030 and plans to expand AI voice ordering beyond 267 drive-thrus.
Hi Auto AI Order Taker expansion · vendor Hi Auto
BCBSA rolls out Sierra voice agent for Federal Employee Program calls
This is a named enterprise deployment at substantial member-service scale; public evidence is currently limited to Sierra’s official announcement and leaves implementation and outcome details open.
Also ConnectOrchestrate
Details2 sources · 1 X post
The Blue Cross Blue Shield Association rolled out a Sierra AI agent for Federal Employee Program member-service calls serving nearly six million people.
Federal Employee Program member-service AI agent · vendor Sierra
SuperDial redesigns healthcare calling for payer-provider agent conversations
Agent-to-agent call handling is a distinct healthcare telephony problem: faster speech is not enough if both endpoints continue exchanging redundant prompts; the 10-million-call figure is company-reported.
Also ConnectOrchestrate
Details2 sources
SuperDial says its healthcare voice-agent platform has passed 10 million provider-payer calls and is redesigning the system for calls where AI agents may sit on both ends.
SuperDial agent-to-agent healthcare calls
PerfectServe uses Twilio for natural voice conversations in Intelligent Answering
This adds a named healthcare operator and use-case to the Twilio voice stack, while public material does not yet provide customer outcomes or a detailed model specification.
Also ConnectOrchestrate
Details2 sources
PerfectServe announced a Twilio-powered natural voice conversation capability in its Intelligent Answering healthcare service.
Intelligent Answering natural voice conversations · vendor Twilio
AteraPlatformlaunch2026-11-17net-new
Atera launches Robin Voice for phone-based IT issue resolution
Voice turns the internal service desk into an inbound support surface for employees; access, escalation, and production outcomes remain to be verified.
Also ConnectOrchestrateThink
Details2 sources
Atera launched Robin Voice, an AI IT-support assistant that lets employees resolve qualifying issues through a phone call.
Robin Voice
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
Talkdesk launches a healthcare platform for AI-assisted patient-services calls
This is more than vertical packaging because the release describes concrete in-call AI and a deployed pharmaceutical workflow; the customer remains unnamed and the outcome claims are vendor-reported.
Also OrchestrateConnectThink
Details2 sources · 1 X post
Talkdesk launched Life Sciences Experience Cloud to connect AI agents, people, systems, and workflows across specialty-care patient services.
Talkdesk Life Sciences Experience Cloud
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