More Reliable Dictaphone Transcription and Smarter Processing

🔄 Improvements

• Improved handling of audio segments that were not transcribed during live recording, helping the system continue processing missing content automatically.
• Enhanced transcription flow stability during network interruptions and temporary transcription service issues.
• Improved processing of audio chunks with better timing alignment, resulting in more accurate transcript positioning.
• Reorganized Dictaphone settings:
  • Buffered Time Window is now easier to find in Standard settings.
  • Paragraph Size has been moved to Advanced settings to reduce clutter.

✨ New Features

• Added automatic processing of previously untranscribed audio segments to help recover missed transcript content.
• Added improved recovery mechanisms for saved audio chunks when transcription results arrive later than expected.
• Added support for importing files from binary blobs, improving compatibility with file import workflows.

🛠 Fixes

• Fixed: Missing transcript text could be lost when Deepgram returned incomplete or empty responses. Buffered transcript recovery now continues correctly.
• Fixed: Audio chunk recovery now continues even when some transcription responses contain no recognized text.
• Fixed: Segmentation processing could incorrectly stop handling sections that had already been promoted for processing.
• Fixed: Several edge cases that could interrupt transcription processing when audio files, recorder instances, or transcription data were unavailable.
• Fixed: Improved reliability of transcript buffering and processing state management during long recordings.

And other improvements.