New feature: ASR automatic model detection

Now you can select multiple ASR models for transcription, and IV will switch intelligently between them within a single recording.  This allows you to:

• transcribe recordings containing speech in multiple languages (see Which languages can IV transcribe?)
• transcribe without being sure of the model in advance
• transcribe with better accuracy recordings where different models are better suited to different speakers, such as a recording containing British and US English speakers.

Improvements & Bug fixes

• New: JumpToWeb export options including specifying a custom range of items to export, and excluding failed items.
• New: JumpToWeb export can now select up to two formats from this expanded list:
  • Alternative Words Lattice (Hyphenated or Probability scored)
  • Audio file
  • Email (with .wav, .mp3, or SmartTranscript attached)
  • SmartTranscript
  • SRT with probabilities
  • Text (plain or US Federal Court format)
  • Word frequency
  • XML
• New: better installation process for models
• Fixed: INT-2878 high MySQL CPU use when re-indexing large groups
• Fixed: INT-2690 exporting from large groups
• Fixed: INT-2882 fixed retention policy handling of already deleted items

Download

Superseded by IV 4.7.1

Documents

No update to the user guides yet, see the 4.6 release page for these.