v0.9.2

What's Changed

Added

• Persistent player: Media player survives page navigation, preserving playback state across routes
• Virtualized transcript list for smooth rendering of long transcripts

Changed

• Transcript controls: Action buttons merged into header; individual export buttons replaced with a dropdown menu
• Storage cleanup: Transcript records deleted atomically alongside media files using a shared helper; clearAllMediaFiles now clears both stores in one transaction; orphaned transcript records pruned during storage cleanup

Fixed

• AI provider and model selections now initialize correctly from localStorage on first render
• Removed obsolete v3 migration block that incorrectly wiped mediaTranscripts on store rehydration

Full Changelog: v0.9.1...v0.9.2

v0.9.1

What's New

Local AI services (offline mode)

• Ollama — use a locally running Ollama instance as the AI provider for chat and language explanation, no cloud API key required
• Local Whisper — connect to a local Whisper server for fully offline audio transcription

Settings overhaul

• Replaced the settings drawer with a unified /settings page
• Added A/B loop controls directly to the player control bar
• Added ?tab=ai deep-link to open the AI settings tab directly

AI explanation UX

• Inline explanation panel replaces the full-screen drawer
• Tighter prompts produce shorter, language-learning-focused output

Transcript & playback controls

• Streamlined transcript action buttons and range-aware transcription entry points
• Transcript empty-state copy now references the active provider
• Translation updates across English, Japanese, and Chinese

AI model catalog

• Refreshed OpenAI, Gemini, and Grok model catalogs with updated pricing metadata
• Updated default model selections to current provider defaults
• Migrated legacy model IDs

Playlist removal

• Removed playlist/queue controls from desktop and mobile UIs
• Removed queue state and media-end auto-advance from the player store

Bug fixes

• Fixed TypeScript build errors caused by unused variables in TranscriptPanel

v0.9.0

Added

• Playback controls:
  • Added a dedicated "play from start" control in the player UI
  • Added safer pending-play handling so media can begin automatically once the element is ready
• Transcription workflow:
  • Added loop-range transcription so the active A-B selection can be sent directly for transcription
  • Added an explicit full-range transcription action and cancel controls for long-running jobs
• Waveform processing:
  • Added cached waveform analysis utilities for background generation and progressive loading
  • Added large-file waveform loading feedback with analysis progress states

Changed

• AI settings overhaul:
  • Rebuilt the AI settings page into clearer defaults, providers, and transcription sections
  • Improved provider selection, model selection, API key handling, and connection testing flows
  • Expanded translations for the new AI and transcription settings UI
• Shadowing workflow simplification:
  • Simplified shadowing playback and recording to a single-track model
  • Refined recorder state handling, cleanup behavior, and waveform rendering for shadowing sessions
• Waveform experience:
  • Reworked waveform loading and rendering to better handle very large local audio files
  • Improved waveform caching, zoom behavior, and loading transitions for stored media
• Player and media state management:
  • Improved playback state synchronization between the media element and the global store
  • Reduced noisy current-time updates and tightened media cleanup during unmounts
• Transcript actions:
  • Refined transcript action buttons and range-aware transcription entry points

Fixed

• Playback reliability:
  • Fixed cases where play requests could be dropped before media was ready
  • Fixed store playback state drifting out of sync with the actual media element
• Waveform stability:
  • Fixed large audio waveform loading regressions and reduced failures during background analysis
  • Improved recovery and status handling when waveform analysis cannot complete normally
• Shadowing and storage cleanup:
  • Improved cleanup around shadowing recordings and related persisted media state
  • Fixed several edge cases in local media storage and waveform cache updates