Soniox Unveils AudioMind Advanced AI Transcription Model

California startup Soniox has unveiled AudioMind, its first AI model capable of deeply understanding audio and transform conversations and transcription into actionable information. Already leading in the English voice-to-text, natural language processing space, Soniox is taking its approach much further by developing AI that is able to fully understand the world and humans through audio, aiming to solve problems requiring real-world interaction.

Ido Gus, Deep Learning Senior Team Leader at CEVA’s Sensor and Audio Business Unit, writes about voice control deployment on low power and resource constrained microcontroller-units (MCUs). The article discusses what to consider for the development of products using voice as a primary human interface, and how speech recognition models using AI technology can support a wide variety of use cases and languages, without compromising on power requirements.

Soniox released a breakthrough in artificial intelligence for speaker diarization (separation) and identification. Soniox Speaker AI has achieved a 96% accuracy in detecting speakers for all kinds of business applications, making it 5x more accurate than solutions provided by Google, Microsoft and Amazon. This breakthrough unlocks a new dimension for voice and conversational applications.