hotwirednews Get GrokBot

Week 27

29 Jun – 5 Jul 20262026-W27 · 16 filings · 13 companies · 2 net-new

The Orchestrate layer led the week with 8 of 16 filings, from Daily / Pipecat, LiveKit, xAI, ElevenLabs and 4 more.

Hearspeech-to-text1

Daily / PipecatPlatformlaunch2026-07-04

Pipecat 1.5.0 adds Together STT and AssemblyAI U3 Pro mode

Together becomes a one-vendor STT/TTS option beside Deepgram and Cartesia pairs.

Details3 sources

pipecat-ai 1.5.0 adds TogetherSTTService and a mode streaming parameter on AssemblyAISTTService for U3 Pro latency profiles.

pipecat-ai 1.5.0

Speaktext-to-speech3

Daily / PipecatPlatformlaunch2026-07-04

Pipecat 1.5.0 adds Together TTS, NVIDIA per-sentence mode, and TTFA metrics

TTFA makes speak-layer latency comparable across providers in the same observer stream.

Details3 sources

pipecat-ai 1.5.0 adds TogetherTTSService, per-sentence synthesis and zero-shot prompts on NvidiaTTSService, and Time To First Audio metrics on TTS services.

pipecat-ai 1.5.0

SonioxModel providermodel release2026-07-02

Soniox launches Voice Cloning across 60-plus languages

Multilingual cloning that explicitly covers names, IDs, and other domain terms targets production voice agents beyond English demos, while the launch leaves sample length and quality benchmarks unspecified.

Details2 sources · 1 X post

Soniox introduced a voice-cloning product that creates a high-fidelity digital voice from a short clear recording and supports more than 60 languages.

Soniox Voice Cloning

Introducing Soniox Voice Cloning. Create a high-fidelity digital replica of a voice from a short clear recording. Soniox Voice Cloning captures the vocal details that make every speaker unique: tone, rhythm, accent, pacing, pronunciation, emotion, delivery, and vocal personality. Built for production use cases: • Voice cloning in 60+ languages, not just English. • Real-world domains, including names, numbers, IDs, addresses, acronyms, product terms, and specialized vocabulary. • Production scale, with pricing for large-volume applications and end-user products. From voice agents to podcasts, audiobooks, ads, games, localization, and enterprise workflows, Soniox Voice Cloning makes it possible to create natural cloned speech that sounds realistic, expressive, and consistent. Try Soniox Voice Cloning today. https://t.co/kiB2MI73Ax

https://x.com/soniox_ai/status/2072715589357834602
TelnyxPlatformmodel release2026-06-30

Telnyx launches Bayan TTS and Basira STT for Arabic voice agents

Telnyx covers both speech directions in one Arabic release, and the 100-plus voices across 13 accents give the TTS half a concrete catalogue claim while Basira’s recognition metrics remain unpublished.

Also HearConnect

Details2 sources · 1 X post

Telnyx launched Bayan for Arabic speech generation and Basira for Arabic speech recognition, with Bayan offering more than 100 voices across 13 Arabic accents.

Bayan TTS and Basira STT

Arabic voice AI has had a weird gap for a while. You could build the agent logic, connect the telephony and get the workflow into production. But the actual speech layer still felt like it was being borrowed from English-first systems. That matters a lot more in Arabic than people think. Arabic is not one clean checkbox. Dialect, accent, pronunciation, pacing, and transcription quality all change how natural the agent feels. Today we’re launching 2 new Telnyx models for solving that problem: Bayan TTS for Arabic speech generation. Basira STT for Arabic speech recognition. Bayan is now available with 100+ Arabic voices across 13 Arabic accents, so teams can generate speech that sounds closer to the markets they actually serve. Basira handles the other side of the conversation: turning spoken Arabic back into text so the agent can understand, route, and respond. Bayan and Basira are our next step toward making Arabic voice agents feel native from both sides of the conversation.

https://x.com/telnyx/status/2072032795300655309

ThinkLLMs, speech-to-speech1

DeepgramModel providermodel release2026-07-01

Deepgram adds Claude Sonnet 5 as a managed Voice Agent API model

Voice-agent builders can change the reasoning model inside Deepgram's managed agent path without wiring a separate Anthropic service; this is a model-provider addition, not a new Deepgram speech model.

Details2 sources · 1 X post

Deepgram made Claude Sonnet 5 available as a managed Anthropic model in its Voice Agent API's Advanced tier.

Claude Sonnet 5 in Voice Agent API · vendor Anthropic

@Claude Sonnet 5 is now available in the Voice Agent API.

https://x.com/DeepgramAI/status/2076762378193031378

Orchestrateframeworks, evals8

Daily / PipecatPlatformlaunch2026-07-04

Pipecat 1.5.0 folds Pipecat Flows into core

Flow-based bots share the same release train and import path as the rest of Pipecat.

Details3 sources

pipecat-ai 1.5.0 brings Pipecat Flows into the core package so conversation graphs install without a separate flows distribution.

pipecat-ai 1.5.0

LiveKitPlatformlaunch2026-07-02

LiveKit hosts Gemma 4 31B inference tuned for voice agents

LiveKit's own latency-first LLM SKU for Agents — 192 ms TTFT vendor claim versus multi-hundred-millisecond hosted peers on the same voice-shaped prompts.

Also Think

Details1 source

LiveKit Inference began serving Gemma 4 31B with voice-shaped serving (long system prompts, large tool catalogs, low queueing) billed on the same LiveKit API key as Agents.

Gemma 4 31B on LiveKit Inference

xAIPlatformlaunch2026-07-01

xAI opens Voice Agent Builder beta on Grok Voice

Packages Grok Voice S2S with telephony and tools under one meter — a platform SKU ahead of Think Fast 2.0 (29 Jul), not another raw model card.

Also ThinkConnectUse

Details2 sources

xAI launched Voice Agent Builder in beta: a no-code console that turns a plain-language call flow, documents, tools, and guardrails into a production phone agent on Grok Voice speech-to-speech.

Voice Agent Builder

ElevenLabsPlatformlaunch2026-07-01

ElevenLabs adds Singapore data residency for enterprise voice products

Local storage and inference remove a deployment blocker for Singapore-serving enterprises across agents, creative, and API products; availability is enterprise-only rather than a general regional rollout.

Also SpeakThink

Details2 sources · 1 X post

ElevenLabs made Singapore data residency available to enterprise customers using ElevenAgents, ElevenCreative, and ElevenAPI, keeping data and inference within Singapore.

Singapore data residency for ElevenAgents, ElevenCreative, and ElevenAPI

Singapore data residency is now available for enterprises. Build and scale with ElevenAgents, ElevenCreative, and ElevenAPI while keeping data and inference within Singapore. https://t.co/cMamnAhwsr

https://x.com/ElevenLabs/status/2072205367728021891
AgoraPlatformlaunch2026-07-01

Agora launches a typed Agents SDK for production voice agents

A single SDK covers three languages and the auth/session layer that production voice agents need, extending Agora’s March Conversational AI Agents offering without claiming a new model or benchmark.

Also ConnectThink

Details2 sources · 1 X post

Agora released a typed Agents SDK for Python, Node.js, and Go that adds authentication, token generation, and session management on its Conversational AI Engine.

Agora Agents SDK

𝗜𝗻𝘁𝗿𝗼𝗱𝘂𝗰𝗶𝗻𝗴 𝗔𝗴𝗼𝗿𝗮 𝗔𝗴𝗲𝗻𝘁𝘀 𝗦𝗗𝗞. Building a voice agent is easy in a demo. Production is where the plumbing shows up: auth, RTC channels, STT/LLM/TTS config, session lifecycle, recovery. Agora Agents SDK helps developers handle that layer faster. https://t.co/5Gy8bExU4E #AIAgents #VoiceAI

https://x.com/AgoraIO/status/2074218422930649373
VercelPlatformAPI2026-06-29net-new

Vercel AI Gateway adds realtime voice, TTS, and transcription

Puts OpenAI-class realtime voice behind the same Gateway metering Next.js teams already use for text — beta on 29 Jun, before OpenAI's 6 Jul gpt-realtime-2.1 API bump.

Also HearSpeakThink

Details1 source

Vercel put realtime voice agents, text-to-speech, and speech-to-text on AI Gateway in beta through AI SDK 7, with the same observability, spend controls, and bring-your-own-key path as its other modalities.

AI Gateway realtime voice

SpeechifyPlatformlaunch2026-06-29

Speechify opens a developer and enterprise platform for APIs and voice agents

Speechify puts its voice APIs and agent-building surface under one developer-facing brand, with a vendor-claimed one-tenth cost position but no named comparison set in the launch post.

Also SpeakThink

Details2 sources · 1 X post

Speechify launched SpeechifyAI for Developers and Enterprises as a single home for its APIs, voice-agent building tools, and documentation.

SpeechifyAI for Developers and Enterprises

Today, we're launching https://t.co/mSVZ3VV63K, the new home for SpeechifyAI for Developers and Enterprises. One platform to access our APIs, build voice agents in minutes, and explore the docs. Or just share an API key with Claude and start vibe building. One of the core innovations behind Speechify has been finding the right balance of cost, quality, and latency for 60M+ users. Now anyone can tap into those same models at one-tenth the cost of existing providers without compromising on speed or how realistic the voices sound. Excited to see what you all build. – Written with @Speechify Voice Typing

https://x.com/SpeechifyAI/status/2071646354368254434
Retell AIPlatformlaunch2026-06-29

Retell launches Conductor, a graph-native copilot for voice agents

Moves Retell from no-code build into continuous ops: failed-call → simulation → reviewable patch, with human approval on every live-agent edit.

Details3 sources

Retell announced Conductor, a review-first AI copilot that proposes edits inside a production voice-agent workflow, generates simulation tests from failed calls, and ships nothing until a human accepts each change.

Conductor

Connecttelephony, channels1

Daily / PipecatPlatformAPI2026-07-04

Pipecat 1.5.0 adds Twilio serializer base_url and RTVI dtmf client message

Browser and mobile clients can dial DTMF without a telephony-only serializer path.

Details3 sources

pipecat-ai 1.5.0 adds base_url on TwilioFrameSerializer and a first-class RTVI dtmf client message for keypad input.

pipecat-ai 1.5.0

Useagents in market2

EvaSpeaksPlatformlaunch2026-07-03net-new

EvaSpeaks introduces an AI voice agent for automotive dealerships

The dated automotive launch establishes EvaSpeaks’ initial vertical wedge before the October omnichannel platform, but the July post gives no customer evidence or rollout detail beyond six-language support.

Also Connect

Details2 sources · 1 X post

EvaSpeaks introduced a dealership voice agent that answers calls, understands context, books appointments, routes intelligently, and supports six languages.

Automotive dealership AI voice agent

Introducing Evaspeaks. An AI voice agent built for automotive dealerships. → Answers every call 24/7 → Understands context → Books appointments → Routes intelligently → 6 languages No missed calls. No voicemail limbo. Just conversations that move forward. #EvaspeaksAI

https://x.com/evaspeaksai/status/2072879503756448231
NTT Docomo BusinessEnterprisepartnership2026-06-29

NTT Docomo Business and ElevenLabs team up on natural-voice contact-centre agents, starting with a Docomo Finance pilot

With NTT Docomo Ventures already an investor, the group's enterprise arm becomes a channel for ElevenAgents into Japanese contact centres, two months after launching its own AI-IVR (2026-04-17-ntt-docomo-business-ancar-routing-ai-ivr, W16); the Docomo Finance pilot has no go-live date. Kotoba's SoftBank distribution (2026-04-17-kotoba-softbank-daredemo-ai, W16) is the nearest carrier pattern with a different vendor (analytical).

Also Speak

Details3 sources

The companies will develop use cases and industry solutions and build ElevenAgents into NTT Docomo Business's services, with a phone-answering proof of concept already running in Docomo Finance's contact centre.

ElevenLabs contact-centre collaboration · vendor ElevenLabs