Voice AI News — Jul 6 – Jul 13, 2026
OpenAI's screenless device & GPT-Live, Gradium's $100M seed, Amazon Moonraker, Nvidia Audex, xAI's 21 Grok voices, and more voice AI this week.
OpenAI Explores Screenless Voice Computing Device
OpenAI is reportedly planning a screenless personal computing device designed by Jony Ive that users will interact with primarily using their voice.
OpenAI Launches Realtime GPT-Live Voice Model
OpenAI has debuted GPT-Live, a conversational voice model designed to make ChatGPT Voice interactions feel more natural by allowing users to interrupt the AI.
Amazon Develops Costly Multi-Action Alexa Agent Project
Internal documents reveal Amazon's costly Moonraker project, which aims to push Alexa into the AI agent race by enabling the assistant to complete multiple actions from a single voice request.
Alexa Plus Offers Natural Voice Conversations
Amazon's paid Alexa+ subscription is launching with features like more natural conversations and improved Fire TV integration.
Google Gemini Expands Live Voice Chats
Google Gemini has expanded into a multi-tool platform offering live voice chats and workspace integrations.
Google Releases Offline Speech To Text App
Google has launched Google AI Edge Eloquent, an offline iOS app that leverages Gemma AI to turn speech into text in real time.
Chrome Beta Improves Voice Dictation Naturalness
Google's Chrome 151 Beta has updated its voice dictation feature to better understand unspoken breaks and punctuation.
Android Phones To Flag Fake Calls
Google is adding a fake call detection feature to Android designed to identify scammers spoofing trusted contacts using cloned voices.
Google Updates Voice Data Training Settings
A new search history setting controls whether Google Lens images, voice search audio, and AI interactions are saved for training.
Samsung Upgrades Voice Recorder With AI
Samsung is introducing cloud-based AI transcription to its native Voice Recorder app in the One UI 9 beta to improve transcript accuracy.
xAI Releases Twenty One Grok Voices
xAI has launched 21 new multilingual voices for Grok Voice alongside improvements to its original five voices.
Nvidia Releases Audex Audio-Text LLM
Nvidia has released Audex, a unified 30-billion-parameter audio-text large language model that understands and generates both audio and speech.
Nissan Launches Tekton With Gemini Voice
The new vehicle model integrates Google's Gemini Voice Assistant to power its in-car digital experience.
New Siri AI Limited To Newer iPhones
Apple confirmed that its upcoming Siri AI launching with iOS 27 this fall will only be supported on iPhone 15 Pro and newer models.
Alibaba Launches CosyVoice Input Method App
Alibaba's Qwen team has launched the CosyVoice input method app, featuring high-accuracy English recognition and intelligent text formatting.
Alibaba Upgrades Real Time Speech Model
Alibaba's Qwen team upgraded its Fun-ASR-Realtime model to support 16 dialects and 30 languages with low latency.
Hollywood Bowl Implements AI Vocal Isolation
The Hollywood Bowl has debuted a major sound system upgrade that uses AI to isolate onstage vocals and reduce noise.
Canada Warns of AI Voice Cloning
Canada's banking regulator has flagged AI voice cloning as an escalating fraud threat capable of bypassing traditional bank security.
Australia Reviews Voice Copyright Protection Laws
Legal experts and residents are examining Australian copyright laws regarding unauthorized AI voice cloning.
German Cities Navigate New Transcription Rules
Municipalities and corporations are testing AI meeting minutes while balancing the privacy implications of Microsoft's updated transcription rules.
Voice AI Startup Gradium Raises $100 Million Seed
Paris-based startup Gradium secured a $100 million seed round backed by Nvidia to develop ultra-low latency audio models that enable near-instantaneous voice agent responses.
Nurix AI Acquires Conversational Startup Verloop
Mukesh Bansal's Nurix AI has acquired conversational AI company Verloop.io for an undisclosed amount to expand its enterprise voice and chat platform.
Ambient Scribes Feasible for ICU Rounds
A pilot study demonstrated the feasibility of using AI-assisted ambient scribes to capture and synthesize progress notes from intensive care unit rounds.
CaringAI Shares Telephone Cognitive Screening Research
Clinical studies show strong agreement between clinicians and CaringAI's voice agent platform for telephone-based cognitive assessments.
AI Algorithm Restores Speech For ALS Patient
A groundbreaking AI algorithm has helped an ALS patient regain his voice by translating brain signals into speech with 97 percent accuracy.
DeepHealth Launches Reporting Pro Radiology Tool
RadNet's DeepHealth subsidiary has commercially launched Reporting Pro, an AI-powered radiology tool that integrates speech recognition.
Voice AI Automates Revenue Cycle Management
SuperDial's CEO envisions a future where most healthcare revenue cycle management is handled by agentic voice AI.
Study Highlights Risks In Clinical STT
A study by a University of Cincinnati researcher reveals critical accuracy risks in clinical speech-to-text AI, emphasizing the ongoing need for human oversight.
Brain-to-Voice BCI Decodes Intelligible Speech
Researchers introduced brain2voice 2.0, a Transformer-based brain-computer interface capable of decoding highly intelligible voice from neural signals in real-time.
ABNB Credit Union Deploys Voice Assistant
The financial institution successfully launched a new Eltropy-powered AI voice digital assistant to improve call efficiency for members.
Automated Calls Drive Massive Loan Generation
Sanjiv Bajaj reported that automated bot calls and AI chats generated Rs 5520 crore in loans at Bajaj Finance in FY26.
Viova Launches AI Call Handling Solution
Viova has rolled out its AI call handling solution for automotive dealers in the UK following international deployments.
ExplorUS Deploys Roommaster AI Voice Concierge
ExplorUS is expanding its partnership with roommaster to deploy an AI voice concierge across more than 20 national and state park lodging sites.
Ookla Warns 5G Networks Unprepared For AI
Ookla reports that most 5G networks lack the uplink speeds and low latency required to support conversational agentic AI.
JAXA Partners to Build Conversational Spacecraft
The Japan Aerospace Exploration Agency has partnered with Idina to develop "Mission Buddy," a voice AI interface for spacecraft.
Critical Security Flaw Found in Dialogflow
Varonis researchers discovered a vulnerability in Google's Dialogflow CX platform that could allow attackers to silently monitor and impersonate customer service chatbots and voice assistants.
Omilia Releases Lexis Generative TTS Model
Omilia has launched Lexis, a generative text-to-speech model designed specifically for enterprise customer experience applications.
Speechify Simba Takes Top Benchmark Spot
Speechify's Simba 3.2 has achieved the number-one spot on two independent voice AI benchmarks, outperforming competitors on quality, speed, and price.
Cohere Releases Open Source Arabic Transcriber
Cohere has launched Transcribe Arabic, an open-source speech recognition model designed to outperform Whisper on regional dialects.
Protoface Integrates Pipecat For RealTime Avatars
Protoface has launched an integration with Pipecat to allow developers to add synchronized, photorealistic avatar video to real-time voice agents.
Springbase Launches Multilingual VoiceToAgent Capabilities
Springbase.ai has introduced a voice assistant supporting over 150 languages that can execute multi-step workflows from spoken commands.
ConvoZen Launches Arabic Speech Models
ConvoZen.AI has introduced Alif and Rawi, two proprietary speech models designed to handle the multi-dialect and code-switched realities of Arabic speakers.
Voqal AI Builds Arabic Voice Technology
Voqal AI is developing voice technology specifically designed to understand the unique way Arabic speakers communicate.
VoicePing Releases Multilingual Enterprise Communication Platform
Version 3.0 of the platform integrates real-time translation, AI meeting minutes, and voice output for corporate use.
Symend Launches Conversational Collections AI Agent
Symend has introduced SymendConverse, an agentic voice AI designed to personalize inbound and outbound collections calls.
Savi Security Launches Voice Protection App
Savi Security has launched an app designed to protect consumers from AI voice cloning scams and ransom calls.
RealityChek Launches Real Time Audio Detection
Hydaway Digital's RealityChek platform has introduced a real-time streaming audio deepfake identification tool.
Reality Defender Warns Of Agentic Callers
Reality Defender has explored a new threat where AI-powered callers mimic legitimate customers to consume enterprise support capacity and bypass fraud metrics.
VoiceType Dictation App Launches Lifetime Subscription
VoiceType has launched a lifetime subscription for $40, offering voice dictation with up to 99.7 percent accuracy at 360 words per minute.
Voiskey Launches Expression Intelligence Typing App
Voiskey has officially launched its voice typing app, which uses a proprietary model to transform spontaneous verbal speech into context-aware text.
CallTower Partners With SESTEK For CX
CallTower has expanded its AI-for-CX portfolio through a strategic partnership with conversational AI provider SESTEK.
AI Studios Launches Roleplay Training Tool
DeepBrain AI's AI STUDIOS has launched AI Roleplay, a conversational training tool for insurance agents to practice customer interactions.
Alphatron And G2 Speech Extend Partnership
Alphatron Zorgverbinders and G2 Speech have extended their strategic partnership to co-develop integrated, AI-powered reporting within the JiveX PACS platform.
Ultrasound Probe Enables Private Silent Speech
Researchers are exploring ultrasound technology to allow users to speak privately and silently with computers.
InnoCaption Surpasses Thirty Million Captioned Calls
InnoCaption celebrated its tenth anniversary and surpassed 30 million captioned phone calls while announcing a leadership transition.
Blueshirt Media Joins CTM Partner Program
Blueshirt Media has joined the CTM Referral Partner Program to pair call tracking with AI answering services for behavioral health organizations.
Togo Launches Fifty Language AI Project
Togo has partnered with Zindi and Togo AI Lab to build AI speech and language models for 50 languages.
Ghana Lab Showcases Inclusive Speech Project
The Responsible AI Lab showcased its inclusive speech recognition project for low-resource languages at the AI for Good Summit.
GPT-Realtime Powers Synthetic Voice Agent Testing
Researchers have implemented a testing system that uses OpenAI's GPT-Realtime 2 model as a synthetic caller to test hosted voice agents.
NigBench Evaluates Multimodal Medical Query Performance
A Nigerian clinical benchmark reveals that large language models perform significantly worse on local-language speech inputs compared to English text.
Hybrid Postprocessing Improves Code-Switched Medical Speech Recognition
A new study evaluated a hybrid approach combining dictionary normalization with large language model postprocessing to enhance automatic speech recognition accuracy for Korean-English code-switched clinical documentation.
Multimodal AI System Automates Arabic Cognitive Screening
Researchers developed a hybrid cognitive screening system that integrates speech processing, computer vision, and large language models to digitize and automate the Arabic version of the Montreal Cognitive Assessment.
Personalized Communication Datasets Developed for Dysarthria Patients
A new protocol outlines the development of high-quality audio and visual datasets to enable personalized, real-time communication technology for children with moderate-to-severe dysarthria.
Somali Speech Recognition Models Evaluated in Benchmarking Study
A controlled comparative study evaluated four end-to-end automatic speech recognition models, including Wav2Vec2.0 and Whisper-small, on a 10-hour Somali speech dataset to address recognition challenges in low-resource languages.
Robust Defense Protects Against Voice Cloning
Researchers developed an identity-entangled, expectation-over-transformation protector that embeds perturbations in audio to disrupt unauthorized zero-shot voice cloning.
Joint Defense Combines Watermarking and Anti-Cloning
A new framework optimizes a single inference-time audio signal to simultaneously support provenance watermarking and prevent downstream voice cloning.
ASR Leaderboard Climbed Via Fine Tuning
Developers climbed 10 places on the Hugging Face Open ASR Leaderboard by fine-tuning their model on the AMI testset.
Sign Language Recognition and Voice Conversion System Developed
Researchers proposed a vision-based sign-to-voice conversion system that uses convolutional neural networks and transformer-based classification to translate hand gestures into spoken voice.
Smart Spectacles Assist Visually Impaired Navigation
Researchers developed wearable smart spectacles integrating offline text-to-speech for real-time indoor obstacle navigation.
Acoustic Simulation Crucial For Physical AI
Engineers are turning to physics-based acoustic simulation to solve far-field speech recognition challenges in physical AI.
Multimodal Framework Analyzes Mood Swings Accurately
Researchers have developed an AI framework that integrates text sentiment, facial expressions, and Mel-spectrogram-based voice analysis to improve emotion recognition.
Brain-Inspired Spiking Neural Network Enhances Speech Recognition
Researchers proposed a task-state fMRI-constrained spiking neural network framework that utilizes human brain functional topology to improve the biological interpretability and performance of speech recognition.
Cough Sequences Explored as Passive Biometric Signals
A study investigated using cough sequences from the Coswara dataset as a passive biometric signal for speaker verification in clinical and remote monitoring environments.
Whisper Model Fine-Tuned for Arabic Stuttering Classification
A new study utilizes OpenAI's Whisper automatic speech recognition model to perform binary fluent versus disfluent classification to improve Arabic speech therapy.
LLM Spoken Dialogue System Assesses Speaking
A study compared second language speaking performance in discussions with a human versus an LLM-driven spoken dialogue system, finding minor differences in pronunciation and speech rate.
Large Language Models Evaluated as Spoken Dialogue Interlocutors
A comparative study analyzed the performance of GPT-4o and Claude 3.5 Sonnet as spoken dialogue partners to assess their effectiveness in evaluating human interactional competence.
Empathetic Spoken Dialogue Systems Tested in Language Assessment
A study compared neutral and empathetic spoken dialogue systems to explore their potential for eliciting interactional competence in second language oral assessments.
Debate Explores AI Examiners in Speaking Tests
Applied linguistics researchers are debating whether spoken dialogue systems powered by generative AI can replace human examiners in high-stakes interactive speaking tests.
Mandarin Listeners Show Reduced Song Transformation
A new study reveals that native Mandarin Chinese speakers experience reduced speech-to-song transformation when listening to speech in their native tonal language.
Voice-Based Authentication Framework for Smart Homes
Researchers have proposed a hybrid CNN-LSTM model utilizing acoustic features to secure smart-home environments against voice spoofing attacks.
Google Read Along Improves ESL Reading
A study in Malaysia demonstrates that Google's AI-powered reading app enhances primary ESL learners' reading fluency and comprehension through automatic speech recognition.
Hybrid Convolution-Transformer Framework Enhances Children's Speech Recognition
A study proposed a hybrid framework based on the Vision Transformer architecture to model spectral and temporal patterns in children's emotional speech.
Post-ASR Correction Framework for Chinese Medicine
A new phonetic fuzzy search method with accent routing reduces word error rates in domain-specific Chinese speech recognition.
Voice AI Economy Reaches Agentic Inflection
Conversational voice agents are handling high-stakes enterprise tasks with sub-300-millisecond response times to deliver human-speed interactions.
AI Audio Sector Reaches Early Profitability
The AI audio sector has achieved profitability faster than AI video, driving an $11 billion valuation surge in six months.
AI Notetakers Raise Meeting Privacy Concerns
While AI notetakers offer quick meeting recaps and summaries, professionals are raising concerns about how the recorded data is handled.
Flair Labs Explains Voice Agent Failures
Flair Labs' co-founder discussed why voice AI agents struggle to convert leads in sectors like real estate.
AI Sound Enhancement Becomes Competitive Advantage
Audio files must be wrapped in metadata like transcripts and speaker IDs before large language models can successfully parse and search them.
Post-Mortem Voice Cloning Raises Legal Governance Concerns
A new paper argues that current legal frameworks are unprepared for the rise of generative AI technologies that can turn deceased individuals' recorded voices and data into autonomous synthetic clones.
Japanese Voice Actors Seek Official Protections
Voice actors in Japan are taking official action to protect their voices as AI replication technology advances in the entertainment industry.
World Cup Scams Leverage Voice Cloning
Kaspersky warned that FIFA World Cup scams are increasingly using AI-powered voice cloning and deepfakes to deceive fans.
UK Laws Lag Behind Voice Cloning
Experts warn that UK residents risk having their voices cloned without consent as current legal frameworks fall behind AI capabilities.
Safe Words Help Prevent Voice Scams
Experts recommend establishing family safe words and limiting online voice exposure to combat increasingly realistic AI voice cloning scams.
BBB Warns Of AI Voice Scams
The Better Business Bureau has issued a warning about AI-generated phone calls and voice cloning scams that have contributed to massive financial losses.
WhatsApp Scams Adopt AI Voice Impersonation
Experts warn that "Hi Mom" family impersonation scams have escalated to include AI-generated voice notes.
Scammers Clone Mother's Voice In Cuba
A Cuban woman reported that scammers cloned her mother's voice using artificial intelligence to steal $1,600 during a solar panel purchase.
AI Voice Fraud Hits Costa Rica
A software studio CFO in Costa Rica was targeted by a WhatsApp voice note scam, highlighting a growing trust gap driven by AI voice cloning.