Oracle Cloud Infrastructure Documentation / Release Notes

Speech has added new languages for Live Transcribe

  • Services: Speech
  • Release Date: August 27, 2025
Speech now supports over 90 languages using the WHISPER model in Live Transcribe, including:
  • auto (automatically detect language)
  • ar (Arabic)
  • en (English)
  • es (Spanish)
  • de (German)
  • fr (French)
  • it (Italian)
  • pt (Portuguese)
  • ja (Japanese)
  • sv (Swedish)
  • nl (Dutch)
WHISPER model can be selected by configuring a new parameter modelType . The default model is ORACLE.
Note

Some parameters are only supported by the ORACLE model. These parameters must not be included in the connection payload when using the Whisper model, or the connection fails. Parameters supported by ORACLE model only -
  • partialSilenceThresholdInMs
  • finalSilenceThresholdInMs
  • stabilizePartialResults
  • shouldIgnoreInvalidCustomizations
  • customizations
If using the WHISPER model, don't include the preceding parameters when establishing the WebSocket connection.
Note

Partial results are supported in the ORACLE model only.