Tag with "Speech Recognition"

French AI startup Mistral AI has unveiled Voxtral Transcribe 2, featuring breakthrough real-time transcription capabilities. Their new models slash processing delays to under 0.2 seconds while supporting multiple languages including Chinese, offering developers powerful tools for voice applications at competitive prices.

iFlytek has launched an AI hardware-software integrated solution capable of accurate speech recognition in 90-decibel noise environments. The technology powers new products like smart office books and translation devices, achieving over 97% accuracy in challenging conditions. The company also introduced custom voice replication using its Spark Speech Large Model.

Alibaba has launched Qwen3-Omni, a cutting-edge multimodal AI model capable of processing audio, video, and text simultaneously. The model outperforms competitors in 32 benchmark tests and rivals Google's Gemini 2.5-Pro in speech recognition. Its innovative training approach mimics human perception, paving the way for advanced applications in customer service and content creation.

OpenAI's Evals tool has introduced native audio input and evaluation features, streamlining the testing of speech recognition and generation models. This update eliminates the need for manual transcription, enhancing efficiency and accuracy for developers working on audio applications.

DingTalk and Tongyi Lab have unveiled Fun-ASR, a cutting-edge speech recognition model designed for multi-industry terminology understanding. The model supports enterprise customization and is already integrated into DingTalk's meeting and voice assistant features. Future enhancements will focus on dialect recognition and noise robustness.

The Institute of Computing Technology at the Chinese Academy of Sciences has unveiled Stream-Omni, a multi-modal AI model based on GPT-4o architecture. This innovative system seamlessly integrates text, vision, and speech modalities, offering real-time interaction capabilities and flexible modal combinations. The model demonstrates superior performance in visual understanding and speech-to-text conversion, marking a significant advancement in AI interaction technologies.

Qwen's CoGenAV introduces a multimodal speech recognition model that synchronizes audio and visual data, significantly improving accuracy in noisy environments. The model achieves state-of-the-art performance across multiple benchmarks while reducing training costs.