hotwirednews Get GrokBot

Week 13

23–29 Mar 20262026-W13 · 25 filings · 16 companies · 6 net-new

The Orchestrate layer led the week with 9 of 25 filings, from Daily / Pipecat, Sierra, IBM, Bland AI and 3 more.

Hearspeech-to-text3

Daily / PipecatPlatformlaunch2026-03-27

Pipecat 0.0.108 adds Deepgram Flux on SageMaker and AssemblyAI medical domain

Self-hosted Flux STT and medical domain recognition extend regulated speech deployments.

Details3 sources

pipecat-ai 0.0.108 adds DeepgramFluxSageMakerSTTService and AssemblyAISTTSettings.domain for modes such as medical-v1, plus on_end_of_turn on AssemblyAI STT.

pipecat-ai 0.0.108

CohereModel providermodel release2026-03-26net-new

Cohere releases Transcribe, a 2B open-weights speech recogniser it ranks first in English on the Open ASR Leaderboard

An enterprise LLM vendor entering open ASR at 2B parameters under Apache 2.0 gives builders a 14-language alternative to Whisper Large v3 and NVIDIA's Parakeet for local and self-hosted transcription; the leaderboard rank is Cohere's own launch-day claim.

Details4 sources · 1 X post

Cohere's first audio model transcribes 14 languages under Apache 2.0, with weights on Hugging Face, a rate-limited free tier on its Audio Transcriptions API, and dedicated Model Vault deployments for production.

Cohere Transcribe (cohere-transcribe-03-2026)

Introducing: Cohere Transcribe – a new state-of-the-art in open source speech recognition.

https://x.com/cohere/status/2037159129345614174
AssemblyAIModel providerlaunch2026-03-25

AssemblyAI makes Medical Mode available for Universal-3 Pro transcription

The correction pass targets errors that can propagate into SOAP notes and referral letters, while AssemblyAI says base-model noise handling and latency stay the same and gives no independent clinical accuracy result.

Details1 source · 1 X post

Medical Mode applies a medical-entity correction pass to Universal-3 Pro pre-recorded and Streaming transcription for clinical workflows.

Medical Mode

Medical Mode is now available for clinical workflows.

https://x.com/AssemblyAI/status/2036822463347302652

Speaktext-to-speech4

Daily / PipecatPlatformlaunch2026-03-27

Pipecat 0.0.108 adds Smallest AI and xAI HTTP TTS

Adds Indian-language-friendly Smallest Waves and xAI HTTP TTS to the vendor matrix.

Details3 sources

pipecat-ai 0.0.108 adds SmallestTTSService for Waves Lightning and XAIHttpTTSService for xAI's HTTP TTS API, plus TTSService.on_turn_context_created.

pipecat-ai 0.0.108

SanasPlatformlaunch2026-03-26net-new

Sanas adds real-time speech-to-speech translation across 13+ languages and upgrades Speech Enhancement

Sanas moves from accent clarity into live speech-to-speech translation for Fortune-scale contact centres; the $60M ARR claim is company-supplied and the language list is given only as "more than 13".

Also Hear

Details2 sources

Sanas launched Real-Time Language Translation with automatic language detection and speaker-identity preservation, plus a Speech Enhancement upgrade that reconstructs foreground audio beyond noise cancellation.

Real-Time Language Translation

falPlatformAPI2026-03-24

fal adds Inworld TTS-1.5 Max for realtime expressive speech

Dated fal primary for a production TTS host ship — fills the prior pass skip that treated Inworld-on-fal as an undated third-party listing.

Details1 source

fal hosted Inworld TTS-1.5 Max with sub-~250ms P90 time-to-first-audio, 15-language support, and pricing around $0.01 per minute.

Inworld TTS-1.5 Max on fal · vendor Inworld

Mistral AIModel providermodel release2026-03-23

Mistral ships Voxtral TTS multilingual model at $0.016 per 1k characters

API TTS at $0.016/1k characters with a claimed 70ms model latency and 3s zero-shot voice adaptation — open-weight reference voices are CC BY-NC, so commercial use still needs the API.

Details1 source

Mistral AI released Voxtral TTS, its first text-to-speech model: a 4B-parameter multilingual system covering nine languages with claimed 70ms model latency, API pricing from $0.016 per 1k characters, and open weights for reference voices under CC BY-NC 4.0.

Voxtral TTS

ThinkLLMs, speech-to-speech3

Daily / PipecatPlatformlaunch2026-03-27

Pipecat 0.0.108 adds Sarvam and Novita LLM services

Indian-market Sarvam chat models become selectable beside OpenAI-compatible endpoints.

Details3 sources

pipecat-ai 0.0.108 adds SarvamLLMService (sarvam-30b/105b variants) and NovitaLLMService, plus developer-role messages across LLM adapters.

pipecat-ai 0.0.108

GoogleModel providermodel release2026-03-26net-new

Google ships Gemini 3.1 Flash Live for realtime voice agents

Realtime S2S with a claimed 90.8% ComplexFuncBench Audio score and SynthID on every clip — LiveKit is named among early adopters for agent stacks.

Also Use

Details1 source

Google released Gemini 3.1 Flash Live, its highest-quality realtime audio and voice model, in preview via the Gemini Live API in Google AI Studio, in Gemini Enterprise for Customer Experience, and under Gemini Live and Search Live for consumers.

Gemini 3.1 Flash Live

Daily / PipecatPlatformlaunch2026-03-23

Pipecat 0.0.107 adds OpenAI Responses LLM service

Responses API becomes a first-class Pipecat LLM path ahead of the 1.0 WebSocket default.

Details3 sources

pipecat-ai 0.0.107 adds OpenAIResponsesLLMService for the OpenAI Responses API with streaming text, function calling, and usage metrics.

pipecat-ai 0.0.107

Orchestrateframeworks, evals9

Daily / PipecatPlatformAPI2026-03-29

Pipecat Context Hub 0.0.11 indexes Daily SDK type docs and adds path-based get_doc

Daily SDK dictionaries and enums become searchable beside method signatures, which ties the hub to Daily's transport; the taint controls are local operator switches rather than an upstream scanning service.

Details1 source

Pipecat Context Hub 0.0.11 indexes RST type documentation for the Daily Python SDK, links methods to those types, and lets get_doc fetch a page by path.

Pipecat Context Hub 0.0.11

SierraPlatformacquisition2026-03-27net-new

Sierra acquires Opera Tech to deepen its Japanese enterprise presence

The Opera Tech team gives Sierra a Tokyo base and named Japanese enterprise relationships as the platform expands in Japan; the 27 March announcement discloses no transaction terms or product migration plan.

Details3 sources · 1 X post

Sierra acquired Opera Tech, a Tokyo-based enterprise AI startup whose team brings local relationships and experience supporting Japanese businesses.

Opera Tech acquisition

Daily / PipecatPlatformAPI2026-03-25

Pipecat Context Hub 0.0.10 turns on cross-encoder reranking and adds Daily Python and domain filters

A cross-encoder model becomes part of every default install, and the release notes claim improved quality without publishing figures; v0.0.9 builds on the call graph from 2026-03-18-pipecat-context-hub-0-0-8.

Details2 sources

Pipecat Context Hub 0.0.10 enables cross-encoder reranking by default, indexes .pyi stubs so daily-co/daily-python becomes a default source, and adds a domain filter to search_examples.

Pipecat Context Hub 0.0.10

IBMPlatformpartnership2026-03-25

IBM adds ElevenLabs TTS and STT to watsonx Orchestrate agents

watsonx Orchestrate builders get ElevenLabs TTS/STT with PCI and Zero Retention options across 70 languages — a managed-voice path for IBM's agent platform rather than a new foundation model.

Also SpeakHear

Details3 sources

IBM and ElevenLabs announced an integration that brings ElevenLabs text-to-speech and speech-to-text into IBM watsonx Orchestrate so builders can add multilingual voice to agentic workflows with enterprise security controls.

watsonx Orchestrate × ElevenLabs · vendor ElevenLabs

Bland AIPlatformlaunch2026-03-25

Bland Norm builds phone agents, pathways, and extraction rules from a typed instruction

Most Bland customers had relied on forward-deployed engineers to structure pathways, so Norm is Bland's self-serve route; it leads to the October dashboard rebuild around a single agent (2026-10-13-bland-dashboard-v2) and the 30-second iMessage agents (2026-10-15-bland-imessage-business-agents).

Details3 sources · 1 X post

Norm is an assistant in the Bland console that writes the prompt, persona, pathways, validation conditions, and extraction rules for a phone agent from a plain-language request.

Norm

Just tell Norm what kind of phone agent you want to build and it does it all for you.

https://x.com/usebland/status/2036826916540604559
ServiceNowPlatformlaunch2026-03-24net-new

ServiceNow open-sources EVA to score voice agents on task accuracy and conversational experience together

The agent under test runs on Pipecat, so EVA joins Daily's aiewf-eval (2026-02-02-daily-llm-benchmark-voice-agents, W06) and STT benchmark (2026-02-13-daily-stt-benchmark-voice-agents, W07) as evaluation harnesses on the same framework, this time from an enterprise platform; 'first to jointly score' is ServiceNow's claim.

Details4 sources

EVA runs bot-to-bot spoken conversations against a voice agent built on Pipecat and reports two scores, EVA-A for accuracy and EVA-X for experience, launched with a 50-scenario airline dataset.

EVA (voice agent evaluation)

ElevenLabsPlatformlaunch2026-03-24

ElevenLabs launches Guardrails 2.0 for policy-controlled ElevenAgents

Custom checks can block replies before delivery and redact entities from transcripts, recordings, and webhook payloads; the 24 March launch is alpha rather than general availability, despite the enterprise policy surface.

Details2 sources · 1 X post

ElevenLabs introduced Guardrails 2.0 in ElevenAgents with custom real-time policy checks, pre-built focus, content, and manipulation protections, and post-call redaction.

Guardrails 2.0 in ElevenAgents

Introducing Guardrails 2.0 in ElevenAgents.

https://x.com/ElevenLabs/status/2036472147904774289
Daily / PipecatPlatformAPI2026-03-23

Pipecat 0.0.107 adds SyncParallelPipeline frame order control

Parallel branches can emit in deterministic order for eval and multi-modal sync.

Details3 sources

pipecat-ai 0.0.107 adds frame_order on SyncParallelPipeline so synchronized outputs can follow pipeline order.

pipecat-ai 0.0.107

LiveKitPlatformlaunch2026-03-23

LiveKit Agents 1.5.1 adds MCPToolset and custom observability endpoints

Toolset and observability orchestration after 1.5.0 defaults. Class-2: folded quiet patch(es) 1.5.7 into this row.

Details2 sources

livekit-agents 1.5.1 adds MCPToolset, LIVEKIT_OBSERVABILITY_URL for custom observability endpoints, and enables AGC by default on RoomInput audio.

livekit-agents 1.5.1

Connecttelephony, channels3

VonagePlatformlaunch2026-03-24net-new

Vonage launches native ServiceNow Voice integration for Contact Center

Calls can trigger incident categorisation and Flow Designer subflows and write resolution data into ServiceNow in real time, so agents stay in CSM or ITSM; the 24 March release gives no customer deployment or performance metric.

Details3 sources · 1 X post

Vonage Contact Center now embeds real-time voice, transcription, and workflow automation in ServiceNow Customer Service Management and IT Service Management.

Vonage Contact Center with ServiceNow Voice · vendor ServiceNow

Daily / PipecatPlatformAPI2026-03-24

Pipecat small-webrtc-transport 1.10.0 reports the bot pc_id as its participant id

Browser code written against the Daily participant model now gets the same bot-connected events on SmallWebRTC, and the 64 KB default now caps what a client can send over that data channel.

Details4 sources

@pipecat-ai/small-webrtc-transport 1.10.0 treats the peer connection pc_id as the bot participant id, so onBotConnected and onBotDisconnected fire in browser clients.

@pipecat-ai/small-webrtc-transport 1.10.0

Daily / PipecatPlatformAPI2026-03-23

Pipecat 0.0.107 syncs output images with audio and optional auto-silence

Avatar and image frames stay lip-syncable with TTS audio on the wire.

Details3 sources

pipecat-ai 0.0.107 adds sync_with_audio on OutputImageRawFrame and audio_out_auto_silence on TransportParams.

pipecat-ai 0.0.107

Useagents in market3

eDreams ODIGEOEnterprisepartnership2026-03-26

eDreams ODIGEO runs inbound support calls in five languages on ElevenAgents

A travel seller with 7.8 million Prime members names ElevenLabs behind its January voice launch, and its April figures (90% of inbound inquiries handled, 33% fewer transfers) are eDO's own and unaudited.

Details2 sources

ElevenLabs published eDreams ODIGEO's move from IVR menus to ElevenAgents voice agents that resolve routine travel inquiries and pass complex calls to the right specialist with context.

Agentic voice customer support (ElevenAgents) · vendor ElevenLabs

ServiceNowPlatformlaunch2026-03-23net-new

ServiceNow brings AI Voice Agents into its mobile apps

ServiceNow moves its voice agents from the phone line into an employee app a day before releasing its EVA harness (2026-03-24-servicenow-eva-voice-agent-eval); W32's Hyro integration (2026-08-05-hyro-servicenow) later brings a third-party voice agent onto the same platform.

Details1 source

Now Assist customers can let employees talk to ServiceNow AI Voice Agents inside the ServiceNow mobile app, with live captions during the call and a transcript afterwards.

AI Voice Agents on Mobile

Sarvam AIPlatformpartnership2026-03-23

Sarvam and Razorpay plan voice-first conversational commerce with Swiggy

The proposed Swiggy-on-Indus path connects voice discovery to agentic payments in one flow, but the 23 March posts say it will initially launch and give no general availability date.

Details3 sources · 1 X post

Sarvam and Razorpay are combining Indian-language models, an agentic stack, and payments so customers can discover products, order, and pay by voice, initially with Swiggy on the Indus app.

Voice-first conversational commerce with Razorpay · vendor Razorpay

Excited to partner with @Razorpay to power the next generation of voice-first, conversational commerce.

https://x.com/SarvamAI/status/2036027689358987396