Article · AI translation
IT Home reported on October 22 that iFlytek announced the debut of its SparkAudio 1.0 foundational speech model at this afternoon’s iFlytek Global 1024 Developer Festival.
SparkAudio 1.0 integrates capabilities including machine translation, speech recognition and semantic understanding, significantly improving fundamental speech capabilities.
The model is claimed to “perform exceptionally well” in real-time speaker separation, environmental understanding, audio understanding and emotion recognition, advancing the application of artificial intelligence in the physical world.
Through an analysis of a film clip, the company demonstrated the model’s ability to accurately identify audio content, background sounds and speakers’ genders, as well as its accuracy in judging scenes.
SparkAudio 1.0 is now fully available in iFlytek Input Method 16.0, iFlytek Tingjian, iFlytek Office Notebook and the iFlytek Translator, supporting switching between Chinese and English at any time and real-time speaker identification.
IT Home noted that iFlytek also released a hardware device called the “iFlytek AI Recording Ring,” equipped with a 4 + 1 ring microphone array and Wi-Fi 6 transmission. The company claims an overall recognition accuracy of over 96% and says it takes just 2 minutes to transcribe a 2-hour recording.