Underdog’s Woof puts text-to-speech in a 31.7 MB browser model, or 40.6 MB for two voices on a CPU. Local speech can mean lower latency, offline access and no server request. #LocalAI #SpeechAI https://spaisee.com/article/a-32mb-speech-model-tests-whether-local-ai-can-beat-the-cloud
A 32MB Speech Model Tests Whether Local AI Can Beat the Cloud
Underdog’s 32MB Woof speech model brings text-to-speech offline, but its noncommercial license could limit adoption by businesses seeking cheaper, more private alternatives to cloud AI.
spaisee.com
November 10, 2026 at 4:02 AM
Underdog’s Woof puts text-to-speech in a 31.7 MB browser model, or 40.6 MB for two voices on a CPU. Local speech can mean lower latency, offline access and no server request. #LocalAI #SpeechAI https://spaisee.com/article/a-32mb-speech-model-tests-whether-local-ai-can-beat-the-cloud
Neural networks can extract accent-neutral speech features, then decode them into audio for a designated accent - aimed at conferencing and recordings. Zoom Communications patent grant #US12776299B1 #Zoom #SpeechAI
November 4, 2026 at 4:21 AM
Neural networks can extract accent-neutral speech features, then decode them into audio for a designated accent - aimed at conferencing and recordings. Zoom Communications patent grant #US12776299B1 #Zoom #SpeechAI
Spotlight on David Harwath, who leads the SALT Lab at UT Austin, building speech AI that learns across modalities to help low-resource languages.
aipeoplewiki.com/people/david...
#DavidHarwath #SpeechAI #AI
aipeoplewiki.com/people/david...
#DavidHarwath #SpeechAI #AI
David Harwath Wiki: Assistant Professor at UT Austin | AI People Wiki
David Harwath, Assistant Professor, Department of Computer Science; leads the SALT Lab at…: sourced bio, career timeline, and official site cs.utexas.edu.
aipeoplewiki.com
November 2, 2026 at 4:58 PM
Spotlight on David Harwath, who leads the SALT Lab at UT Austin, building speech AI that learns across modalities to help low-resource languages.
aipeoplewiki.com/people/david...
#DavidHarwath #SpeechAI #AI
aipeoplewiki.com/people/david...
#DavidHarwath #SpeechAI #AI
NVIDIA released Nemotron 3.5 Transcribe, a free open-weight speech model that transcribes 40 languages, translates 31 to or from English and tells up to eight speakers apart 🎙️
huggingface.co/blog/nvidia/...
#NVIDIA #OpenSource #SpeechAI
huggingface.co/blog/nvidia/...
#NVIDIA #OpenSource #SpeechAI
October 29, 2026 at 8:26 PM
NVIDIA released Nemotron 3.5 Transcribe, a free open-weight speech model that transcribes 40 languages, translates 31 to or from English and tells up to eight speakers apart 🎙️
huggingface.co/blog/nvidia/...
#NVIDIA #OpenSource #SpeechAI
huggingface.co/blog/nvidia/...
#NVIDIA #OpenSource #SpeechAI
Applied Brain Research announces the general availability of its #speechAI SDK for Zephyr RTOS on the @alifsemi Ensemble family of #EdgeAIProcessors.
ABR Launches Speech AI SDK for Alif Processors
Applied Brain Research (ABR) has announced the general availability of its speech AI software development kit (SDK) for Zephyr RTOS on the Alif Semiconductor Ensemble family of edge AI processors. The software enables streaming automatic speech recognition and text-to-speech synthesis to run entirely on-device, without requiring a cloud connection. It is designed for low-power embedded applications, including battery-operated and air-gapped products. The SDK combines two speech technologies. Niagara ASR (automatic speech recognition) provides low-latency, full-vocabulary streaming speech recognition, continuously converting spoken input into text as it is received. Nith TTS…
www.embedded.com
November 2, 2026 at 9:25 PM
Applied Brain Research announces the general availability of its #speechAI SDK for Zephyr RTOS on the @alifsemi Ensemble family of #EdgeAIProcessors.
🧠 Classification, detection, or segmentation? Chained pipeline or Voice Live? Studying for AI-901? Learn how AI sees, hears, speaks, and creates images, and which Azure service fits.
👉Watch now! buff.ly/XSz0BbV
#AI901 #AzureAI #GenerativeAI #SpeechAI #MVPBuzz
👉Watch now! buff.ly/XSz0BbV
#AI901 #AzureAI #GenerativeAI #SpeechAI #MVPBuzz
How AI Sees and Hears for AI-901: Hear, Speak, See, and Create - Ciraltos
AI-901 Speech and Computer Vision Explained: How AI Hears, Speaks, Sees, and Creates Images
www.ciraltos.com
October 30, 2026 at 4:04 PM
🧠 Classification, detection, or segmentation? Chained pipeline or Voice Live? Studying for AI-901? Learn how AI sees, hears, speaks, and creates images, and which Azure service fits.
👉Watch now! buff.ly/XSz0BbV
#AI901 #AzureAI #GenerativeAI #SpeechAI #MVPBuzz
👉Watch now! buff.ly/XSz0BbV
#AI901 #AzureAI #GenerativeAI #SpeechAI #MVPBuzz
This AI voice model talks more like a real person.
Meet Dessa a speech AI that responds in real time, handles interruptions, and expresses emotion. No converting speech to text first.
Would you be able to tell the difference?
#VoiceAI #SpeechAI #ConversationalAI
Meet Dessa a speech AI that responds in real time, handles interruptions, and expresses emotion. No converting speech to text first.
Would you be able to tell the difference?
#VoiceAI #SpeechAI #ConversationalAI
October 29, 2026 at 9:29 AM
This AI voice model talks more like a real person.
Meet Dessa a speech AI that responds in real time, handles interruptions, and expresses emotion. No converting speech to text first.
Would you be able to tell the difference?
#VoiceAI #SpeechAI #ConversationalAI
Meet Dessa a speech AI that responds in real time, handles interruptions, and expresses emotion. No converting speech to text first.
Would you be able to tell the difference?
#VoiceAI #SpeechAI #ConversationalAI
Nvidia says Nemotron 3.5 Transcribe ranks first among open-weight models on the Hugging Face English ASR leaderboard at 5.33% WER. Benchmark figures are Nvidia's own. #SpeechAI #Nvidia
https://bymachine.news/nvidia-nemotron-3-5-transcribe-granary-v2
https://bymachine.news/nvidia-nemotron-3-5-transcribe-granary-v2
Nvidia Releases Open Nemotron 3.5 Transcribe for 40 Languages
Nvidia releases open Nemotron 3.5 Transcribe, covering 40 languages, plus Granary v2 dataset with 463,062 hours. Benchmark claims and limits.
bymachine.news
October 29, 2026 at 12:30 PM
Nvidia says Nemotron 3.5 Transcribe ranks first among open-weight models on the Hugging Face English ASR leaderboard at 5.33% WER. Benchmark figures are Nvidia's own. #SpeechAI #Nvidia
https://bymachine.news/nvidia-nemotron-3-5-transcribe-granary-v2
https://bymachine.news/nvidia-nemotron-3-5-transcribe-granary-v2
What changes when AI learns conversation directly from sound?
Block’s Dessa was reportedly trained entirely on human speech. It listens and speaks simultaneously, learning timing, laughter and intonation alongside words.
block.xyz/dessa
#SpeechAI #MultimodalAI #HumanAIInteraction
Block’s Dessa was reportedly trained entirely on human speech. It listens and speaks simultaneously, learning timing, laughter and intonation alongside words.
block.xyz/dessa
#SpeechAI #MultimodalAI #HumanAIInteraction
October 25, 2026 at 7:34 PM
What changes when AI learns conversation directly from sound?
Block’s Dessa was reportedly trained entirely on human speech. It listens and speaks simultaneously, learning timing, laughter and intonation alongside words.
block.xyz/dessa
#SpeechAI #MultimodalAI #HumanAIInteraction
Block’s Dessa was reportedly trained entirely on human speech. It listens and speaks simultaneously, learning timing, laughter and intonation alongside words.
block.xyz/dessa
#SpeechAI #MultimodalAI #HumanAIInteraction
Dessa talks like us—full-duplex voice AI with emotion, sound, and pauses. #VoiceAI #SpeechAI #AI #Block #Dessa #EthicalAI thedailytechfeed.com/block-unveil...
October 24, 2026 at 4:32 PM
Dessa talks like us—full-duplex voice AI with emotion, sound, and pauses. #VoiceAI #SpeechAI #AI #Block #Dessa #EthicalAI thedailytechfeed.com/block-unveil...
Dessa just taught Block to hear the feelings behind words, not just the text. Their new speech‑AI runs full‑duplex on NVIDIA GB200 and even on a MacBook Pro M4. Curious how AI finally gets human emotion? Dive in! #Dessa #SpeechAI #EmotionDetection
🔗 aidailypost.com/news/block-t...
🔗 aidailypost.com/news/block-t...
October 24, 2026 at 2:33 PM
Dessa just taught Block to hear the feelings behind words, not just the text. Their new speech‑AI runs full‑duplex on NVIDIA GB200 and even on a MacBook Pro M4. Curious how AI finally gets human emotion? Dive in! #Dessa #SpeechAI #EmotionDetection
🔗 aidailypost.com/news/block-t...
🔗 aidailypost.com/news/block-t...
Grok Voice 3.0 just took the #1 spot for voice AI, beating GPT and Gemini.
It’s speech to speech, so you talk and it talks back in under a second.
It runs as one model instead of three steps, so it’s faster and actually fixes support problems.
#AI #VoiceAI #Grok #xAI #AInews #SpeechAI #AITools
It’s speech to speech, so you talk and it talks back in under a second.
It runs as one model instead of three steps, so it’s faster and actually fixes support problems.
#AI #VoiceAI #Grok #xAI #AInews #SpeechAI #AITools
October 20, 2026 at 12:42 PM
A new arXiv study interprets WavLM's layers for spoof detection, finding that combining five specific layers outperforms full thirteen-layer pooling for attack-level evaluation. CCA analysis reveals mid-layers best align with features…
#AI #ASV #SpeechAI #WavLM
https://arxiv.org/abs/2610.23526
#AI #ASV #SpeechAI #WavLM
https://arxiv.org/abs/2610.23526
October 19, 2026 at 8:01 PM
A new arXiv study interprets WavLM's layers for spoof detection, finding that combining five specific layers outperforms full thirteen-layer pooling for attack-level evaluation. CCA analysis reveals mid-layers best align with features…
#AI #ASV #SpeechAI #WavLM
https://arxiv.org/abs/2610.23526
#AI #ASV #SpeechAI #WavLM
https://arxiv.org/abs/2610.23526
Bee’s new AI wearables Frontier and Eclipse listen so you don’t have to. #AI #Wearables #AmazonBee #Privacy #Productivity #SpeechAI https://thedailytechfeed.com/amazons-bee-unveils-bee-frontier-bee-eclipse-personal-ai-devices-from-75/
October 15, 2026 at 12:18 PM
Bee’s new AI wearables Frontier and Eclipse listen so you don’t have to. #AI #Wearables #AmazonBee #Privacy #Productivity #SpeechAI https://thedailytechfeed.com/amazons-bee-unveils-bee-frontier-bee-eclipse-personal-ai-devices-from-75/
A new arXiv paper proposes RASSG, a method for streaming speech assistants that gates tool launch and cancellation based on argument stability and calibrated invalidity risk to avoid irreversible side effects from stale…
#speechAI #tooling #streaming #arXiv
https://arxiv.org/abs/2610.14659
#speechAI #tooling #streaming #arXiv
https://arxiv.org/abs/2610.14659
October 15, 2026 at 12:01 PM
A new arXiv paper proposes RASSG, a method for streaming speech assistants that gates tool launch and cancellation based on argument stability and calibrated invalidity risk to avoid irreversible side effects from stale…
#speechAI #tooling #streaming #arXiv
https://arxiv.org/abs/2610.14659
#speechAI #tooling #streaming #arXiv
https://arxiv.org/abs/2610.14659
🎙️ Can phase space dynamics boost ASR?
Most speech systems ignore signal phase. In our paper in SIVP, we use Recurrence Plots + 2D Adaptive Wavelets to capture non-linear dynamics.
📄 Read more: link.springer.com/article/10.1...
#SpeechAI #ASR #SignalProcessing #MachineLearning
Most speech systems ignore signal phase. In our paper in SIVP, we use Recurrence Plots + 2D Adaptive Wavelets to capture non-linear dynamics.
📄 Read more: link.springer.com/article/10.1...
#SpeechAI #ASR #SignalProcessing #MachineLearning
Improvement of automatic speech recognition systems utilizing 2D adaptive wavelet transformation applied to recurrence plot of speech trajectories - Signal, Image and Video Processing
Spectral-based features, typically used in ASR systems, do not capture the phase information of speech signals. Thus, exploiting new features that do not ignore the phase of the signal can be a comple...
link.springer.com
October 6, 2026 at 2:37 AM
🎙️ Can phase space dynamics boost ASR?
Most speech systems ignore signal phase. In our paper in SIVP, we use Recurrence Plots + 2D Adaptive Wavelets to capture non-linear dynamics.
📄 Read more: link.springer.com/article/10.1...
#SpeechAI #ASR #SignalProcessing #MachineLearning
Most speech systems ignore signal phase. In our paper in SIVP, we use Recurrence Plots + 2D Adaptive Wavelets to capture non-linear dynamics.
📄 Read more: link.springer.com/article/10.1...
#SpeechAI #ASR #SignalProcessing #MachineLearning
Suno AI 음성 생성 기능 3가지 활용법 공개
https://bit.ly/4j1gwPs
#SunoAI #AI음성 #인공지능 #SpeechAI #AI기술 #ArtificialIntelligence #VoiceGeneration
https://bit.ly/4j1gwPs
#SunoAI #AI음성 #인공지능 #SpeechAI #AI기술 #ArtificialIntelligence #VoiceGeneration
Suno AI 음성 생성 기능 3가지 활용법 공개 - 천의무봉
AI 음악 생성 서비스로 유명했던 Suno가 이제 말하는 목소리까지 만들어냅니다. 새롭게 공개된 Speech 기능은 텍스트를 입력하면 합성 음성을 생성하고, 여기에 AI 배경음악까지 하나의 트랙으로 붙여주는 방식입니다. 기존의 텍스트 음성 변환 도구와는 결이 조금 다르기 때문에,
blog.hangadac.com
October 2, 2026 at 1:00 PM
Suno AI 음성 생성 기능 3가지 활용법 공개
https://bit.ly/4j1gwPs
#SunoAI #AI음성 #인공지능 #SpeechAI #AI기술 #ArtificialIntelligence #VoiceGeneration
https://bit.ly/4j1gwPs
#SunoAI #AI음성 #인공지능 #SpeechAI #AI기술 #ArtificialIntelligence #VoiceGeneration
Reached ~87% test accuracy and ~93% validation accuracy on spontaneous speech with our proposed CNN framework.
Demonstrates that clinical AI needs both acoustic timbre and phonetic context.
Check out the full research here: link.springer.com/article/10.1...
#SpeechAI #Acoustics #DeepLearning
Demonstrates that clinical AI needs both acoustic timbre and phonetic context.
Check out the full research here: link.springer.com/article/10.1...
#SpeechAI #Acoustics #DeepLearning
Evaluation of phone posterior probabilities for pathology detection in speech data using deep learning models - International Journal of Speech Technology
Voice pathology detection (VPD) aims to accurately identify voice impairments by analyzing speech signals. This study proposes models based on deep learning (DL) for binary classification to distingui...
link.springer.com
September 25, 2026 at 8:15 AM
Reached ~87% test accuracy and ~93% validation accuracy on spontaneous speech with our proposed CNN framework.
Demonstrates that clinical AI needs both acoustic timbre and phonetic context.
Check out the full research here: link.springer.com/article/10.1...
#SpeechAI #Acoustics #DeepLearning
Demonstrates that clinical AI needs both acoustic timbre and phonetic context.
Check out the full research here: link.springer.com/article/10.1...
#SpeechAI #Acoustics #DeepLearning
Low-level acoustic features (MFCCs) only tell half the story in voice pathology detection.
What happens when we introduce phonetic-based features (PPPs) into deep learning models?
🧵 Our paper in IJST (Springer) explores this:
link.springer.com/article/10.1...
#SpeechAI #Acoustics #Phonetics
What happens when we introduce phonetic-based features (PPPs) into deep learning models?
🧵 Our paper in IJST (Springer) explores this:
link.springer.com/article/10.1...
#SpeechAI #Acoustics #Phonetics
Evaluation of phone posterior probabilities for pathology detection in speech data using deep learning models - International Journal of Speech Technology
Voice pathology detection (VPD) aims to accurately identify voice impairments by analyzing speech signals. This study proposes models based on deep learning (DL) for binary classification to distingui...
link.springer.com
September 25, 2026 at 8:13 AM
Low-level acoustic features (MFCCs) only tell half the story in voice pathology detection.
What happens when we introduce phonetic-based features (PPPs) into deep learning models?
🧵 Our paper in IJST (Springer) explores this:
link.springer.com/article/10.1...
#SpeechAI #Acoustics #Phonetics
What happens when we introduce phonetic-based features (PPPs) into deep learning models?
🧵 Our paper in IJST (Springer) explores this:
link.springer.com/article/10.1...
#SpeechAI #Acoustics #Phonetics
Key Numbers:
• ~85% accuracy on challenging test sets
• ~92% accuracy on evaluation sets
Proving that spontaneous speech isn’t just “noise”—it’s rich in diagnostic
Check out the paper here: link.springer.com/article/10.1...
#SpeechAI #AudioProcessing #MachineLearning
• ~85% accuracy on challenging test sets
• ~92% accuracy on evaluation sets
Proving that spontaneous speech isn’t just “noise”—it’s rich in diagnostic
Check out the paper here: link.springer.com/article/10.1...
#SpeechAI #AudioProcessing #MachineLearning
Voice pathology detection on spontaneous speech data using deep learning models - International Journal of Speech Technology
Speech problems are a common issue that affects people everywhere and can affect the quality of their lives. The human speech production system involves various components. Dysfunction of any of these...
link.springer.com
September 25, 2026 at 8:09 AM
Key Numbers:
• ~85% accuracy on challenging test sets
• ~92% accuracy on evaluation sets
Proving that spontaneous speech isn’t just “noise”—it’s rich in diagnostic
Check out the paper here: link.springer.com/article/10.1...
#SpeechAI #AudioProcessing #MachineLearning
• ~85% accuracy on challenging test sets
• ~92% accuracy on evaluation sets
Proving that spontaneous speech isn’t just “noise”—it’s rich in diagnostic
Check out the paper here: link.springer.com/article/10.1...
#SpeechAI #AudioProcessing #MachineLearning
In our new paper in IEEE Access, we use Log-Area Ratios (LARs) + a novel Conditional Speaker Normalization (CSN) conditioned on speaker proxies (e.g. height) to detect synthetic speech reliably on ASVspoof & FoR.
ieeexplore.ieee.org/abstract/doc...
#DeepfakeDetection #SpeechAI #AudioDeepfake
ieeexplore.ieee.org/abstract/doc...
#DeepfakeDetection #SpeechAI #AudioDeepfake
Conditional Speaker Normalization of Vocal Tract Shape Features for Robust Synthetic Speech Detection
The rapid advancement of text-to-speech and voice conversion technologies has significantly improved the quality of synthetic speech, posing increasing challenges for developing reliable detection cou...
ieeexplore.ieee.org
September 20, 2026 at 4:04 AM
In our new paper in IEEE Access, we use Log-Area Ratios (LARs) + a novel Conditional Speaker Normalization (CSN) conditioned on speaker proxies (e.g. height) to detect synthetic speech reliably on ASVspoof & FoR.
ieeexplore.ieee.org/abstract/doc...
#DeepfakeDetection #SpeechAI #AudioDeepfake
ieeexplore.ieee.org/abstract/doc...
#DeepfakeDetection #SpeechAI #AudioDeepfake
)
🎙️ Are synthetic speech artifacts evenly distributed across phonemes? Not quite!
In our latest paper in MTAP, we propose a phonetic-driven framework using PPGs & phoneme pruning, boosting anti-spoofing performance on ASVspoof 2019 LA.
🔗 link.springer.com/article/10.1...
#SpeechAI #AudioDeepfake
🎙️ Are synthetic speech artifacts evenly distributed across phonemes? Not quite!
In our latest paper in MTAP, we propose a phonetic-driven framework using PPGs & phoneme pruning, boosting anti-spoofing performance on ASVspoof 2019 LA.
🔗 link.springer.com/article/10.1...
#SpeechAI #AudioDeepfake
Client Challenge
link.springer.com
September 18, 2026 at 6:21 AM
)
🎙️ Are synthetic speech artifacts evenly distributed across phonemes? Not quite!
In our latest paper in MTAP, we propose a phonetic-driven framework using PPGs & phoneme pruning, boosting anti-spoofing performance on ASVspoof 2019 LA.
🔗 link.springer.com/article/10.1...
#SpeechAI #AudioDeepfake
🎙️ Are synthetic speech artifacts evenly distributed across phonemes? Not quite!
In our latest paper in MTAP, we propose a phonetic-driven framework using PPGs & phoneme pruning, boosting anti-spoofing performance on ASVspoof 2019 LA.
🔗 link.springer.com/article/10.1...
#SpeechAI #AudioDeepfake
Discover Meta Muse Voice Transcribe 1.0, how its real-time speech AI works, key features, performance, pricing and use cases.
blog.9cv9.com/what-is-meta...
#Meta, #MuseVoiceTranscribe, #MuseVoiceTranscribe1, #MetaAI, #MetaAIResearch, #VoiceAI, #SpeechAI,
blog.9cv9.com/what-is-meta...
#Meta, #MuseVoiceTranscribe, #MuseVoiceTranscribe1, #MetaAI, #MetaAIResearch, #VoiceAI, #SpeechAI,
What is Meta: Muse Voice Transcribe 1.0, How Does It Work & Use Cases
Meta Muse Voice Transcribe 1.0 is a real-time speech AI model that combines transcription, speaker diarization, endpoint detection and multilingual code-switching. Discover how it works, its performan...
blog.9cv9.com
September 16, 2026 at 5:50 PM
Discover Meta Muse Voice Transcribe 1.0, how its real-time speech AI works, key features, performance, pricing and use cases.
blog.9cv9.com/what-is-meta...
#Meta, #MuseVoiceTranscribe, #MuseVoiceTranscribe1, #MetaAI, #MetaAIResearch, #VoiceAI, #SpeechAI,
blog.9cv9.com/what-is-meta...
#Meta, #MuseVoiceTranscribe, #MuseVoiceTranscribe1, #MetaAI, #MetaAIResearch, #VoiceAI, #SpeechAI,