More Control Over Transcription: New Segmentation Modes, Smarter Headings, Better Timing

✨ New Features

• Added toggle for AI paragraph formatting, giving users control over whether transcription is grouped by AI or kept raw
  Screenshot
• Added toggle for semantic segmentation, allowing users to switch between simple word/time cuts and meaning-based segmentation
  Screenshot
• Introduced time-based buffer segmentation mode with adjustable seconds ranges
  Screenshot
• Added section size options (seconds) for segmentation, enabling finer control over how long each segment is
• Improved minimap generation with heading-deduplication based on sibling branches, ensuring headings stay unique
• New AI model added: gpt-4.1-nano-2025-04-14 with better logic for selecting main words and producing clear parent-child structures

🔄 Improvements

• Updated note behavior: section notes now always store full text when quote tasks exist, ignoring AI-shortened snippets
• Improved timing accuracy in live transcription by syncing to the actual first audio packet
  Fixes previous drift caused by WebSocket delays
• Enhanced minimap creation by checking previous headings to avoid repetition
• Increased minimap processing limit from 30 to 70 seconds for slower models
• Better timeout handling for slow AI models (up to 60 seconds)
• Better organization of the "New" menu for creating content: Meetings, Media, Text

🛠 Fixes

• Fixed: minimap no longer freezes during live transcription
• Fixed: more accurate timestamp assignment for words during live transcription

And other improvements