AI
Meta’s Omnilingual ASR Recognized by TIME
Meta’s Omnilingual ASR system, which supports over 1,600 languages, has been recognized as one of TIME’s Best Inventions of 2026.
October 16, 2026, 01:35 AM·Crypto Briefing
#speech-recognition
AI
TII Unveils Falcon-ASR Speech Model for Arabic and Multilingual Tasks
TII has launched Falcon-ASR, a 1.6B parameter speech-to-text model supporting Arabic and four other languages, achieving a 20.92% WER in internal tests.
MarkTechPost
AI
Willow Care partners with CoreWeave to scale AI training
Willow Care Inc. has partnered with CoreWeave to manage its AI training infrastructure, aiming to replace traditional keyboard input with voice-based interaction within two years.
SiliconANGLE
AI
Anthropic Launches Voice Data Sharing for AI Model Training
Anthropic has launched a program allowing users to voluntarily share voice data to improve the speech recognition and response capabilities of its Claude AI models.
Crypto Briefing
AI
Microsoft Expands Real-Time Voice AI Portfolio
Microsoft has launched three new AI models for real-time voice processing, including a top-ranked streaming transcription tool and two text-to-speech engines with voice cloning capabilities.
cnBeta.COM
AI
MTS Reports Usage Statistics for Hearing Impairment Service
MTS reported that its service for the hearing impaired processed over 496,000 calls across Russia, including the Sakhalin region, between March and August 2026.
CNews.ru
Money
Klang Secures €1,32 Million to Scale AI Conversation Analysis
Helsingborg-based AI startup Klang has raised €1,32 million to fuel international expansion and research, while simultaneously releasing an open-source speech-to-text model for Swedish.
EU-Startups
AI
MTS Reports Usage Statistics for Its Hearing Impairment Service
MTS reported that its AI-powered service for the hearing impaired has facilitated nearly 500,000 calls across Russia in the last six months.
CNews.ru
AI
iFlytek Introduces Spark-ASR-2.0 Speech Recognition Model
iFlytek has unveiled its Spark-ASR-2.0 speech recognition model, which offers improved accuracy in complex acoustic environments with only a 10 percent increase in inference costs.
IT之家
AI
Kyutai Introduces Voice of Reason for Spoken Math Reasoning
Kyutai has launched Voice of Reason, a speech-native model that applies reinforcement learning to solve spoken math problems, achieving up to 77.1% on the GSM8K benchmark.
MarkTechPost
AI
StepStar Unveils StepAudio 3 Series for Advanced Voice Interaction
StepStar has launched the StepAudio 3 series, featuring five models that lead global benchmarks in real-time interaction, speech recognition, and audio generation.
IT之家
AI
Xiaomi Introduces CocktailASR-1 for Multi-Speaker Speech Recognition
Xiaomi has launched an open-source AI model, CocktailASR-1, capable of isolating and transcribing a specific speaker's voice within multi-speaker audio recordings.
3DNews - Daily Digital Digest
AI
Xiaomi Launches Open-Source CocktailASR-1 Speech Model
Xiaomi has open-sourced its new CocktailASR-1 model, designed to isolate and transcribe a specific speaker's voice in multi-person environments using end-to-end LLM architecture.
IT之家
AI
Kryptonite Engineers Optimize Language Identification Model
Kryptonite researchers have created a language identification model that is 10 times faster and 5 times lighter than the original 1-billion-parameter baseline.
Хабр
AI
Kryptonite Launches Compact Speech Language Identification Model
Kryptonite has developed a compact speech language identification model that is 10 times faster and 5 times smaller than its analog.
CNews.ru
That is everything