Physix Frontier · News Briefing Card (IT Home · Nov 12, 2026)

iFlytek Launches SparkAudio 1.0 Speech Foundation Model API

KEY FACTS

  • iFlytek announced on November 12 that SparkAudio 1.0 is now available as an API service on the iFlytek Open Platform.
  • Developers can directly integrate and call this speech foundation model.
  • The model supports text and speech input with text output, covering tasks such as transcription, translation, and sentiment analysis.
  • The model supports recognition and translation across 99 languages and 202 dialects.
  • The model is trained on fully domestic compute and targets scenarios such as intelligent customer service and automotive cockpits.

KEY DATA

99Supported languages
202Supported dialects

PHYSIX OBSERVATION

A speech foundation model that understands audio directly bypasses the cascaded pipeline of transcribe-then-understand, reducing information loss and latency. For developers, multilingual and multidialect coverage plus fully domestic compute is a differentiating selling point, but real-world performance will depend on API call costs and robustness in actual scenarios.

Source: IT Home report