DualPiP v1.7 Release Notes

This update introduces ASR (Automatic Speech Recognition) live captions, generating real-time AI subtitles for any web video with full Learning Mode and AI translation support. It also adds online subtitle search, local subtitle import, and a completely rebuilt AI translation architecture.

1.7.4 - 2026-06-05

• Auto Live Captions: ASR subtitles now start automatically when no traditional subtitles are detected.
• Added "Auto ASR for this site" toggle in the PiP Live Captions menu for easy per-site management.

• New guided onboarding page for first-time users with interactive step-by-step setup.
• ASR language preference is now remembered per site, prioritizing site-specific language over global default.
• Improved Live Captions matching accuracy for X (Twitter) timeline videos.
• Optimized resource loading strategy, reducing reliance on proxy servers.

• Fixed media resource requests failing on X (Twitter) videos.

1.7.3 - 2026-06-03

• Improved stability of PiP video playback after switching browser tabs.
• Improved rendering performance of the translation service settings page.
• Improved interaction experience of the add model dialog.
• Live captions now follow the subtitle style settings (font, color, size, etc.).
• Added Live Captions shortcut key (Shift+A) display and customization in Options and PiP settings.
• Toggling Live Captions now shows a toast notification with current recognition language.
• Added browser language compatibility hint for the multilingual mode in language list.
• Karaoke Mode renamed and added an "Off" option.
• Added content scope description to the subtitle search panel.
• Improved scrolling behavior for long bilingual Live Captions.

• Fixed subtitle style settings not taking effect on first click.
• Fixed Live Captions overflowing when text wraps to multiple lines.
• Fixed Learning Mode subtitle list not updating highlight correctly during gaps between subtitles.

1.7.2 - 2026-05-31

• Translation service settings now use a tabbed layout for a clearer, more focused experience.
• Adding a model now supports dropdown selection, with the ability to automatically fetch available models from the provider.
• Live captions language support expanded to over 50 languages.
• Live captions preload mode now compatible with more video platforms.

1.7.1 - 2026-05-30

• Improved ASR provider settings with model selection and multi-provider switching.

• Fixed ASR live captions not displaying translations correctly for certain languages.
• Fixed ASR auto mode incorrectly identifying live streams as on-demand videos.

1.7.0 - 2026-05-29

• Deepgram Cloud Recognition: Real-time WebSocket streaming with sub-300ms latency, supporting 22+ languages with automatic detection.
• Local Whisper Recognition: Connect to locally deployed OpenAI-compatible Whisper servers for privacy-friendly, offline speech recognition.
• Streaming + Pre-download Dual Modes: Auto mode tries WebSocket streaming first, falls back to pre-download batch recognition when streaming is unavailable.
• Learning Mode Integration: ASR captions fully support Learning Mode subtitle panel, AB loop repeat, and all learning features.
• AI Translation Bilingual Captions: Combine ASR captions with AI translation to generate real-time bilingual subtitles.
• Keyboard Shortcuts: Press Shift+A in the PiP window to toggle ASR, or configure a global shortcut via chrome://extensions/shortcuts.
• ASR Settings Page: Add, edit, and manage speech recognition providers with language, capture mode, and other configuration options.

• Online Subtitle Search: Search and download subtitles directly from the PiP subtitle menu, supporting DualPiP built-in, OpenSubtitles, and SubDL sources.
• Local Subtitle Import: Import local subtitle files in SRT/VTT/ASS/SSA/SUB formats.
• Smart Identification: Automatically detects video IMDb ID, TMDb ID, and title for precise search results.
• Search Settings: Configure custom API keys, with source switching via dropdown menu that remembers your choice.
• Free users can search subtitles unlimited times, with 2 downloads per day.

• Rebuilt AI translation architecture with separate provider and model management, 30+ preset AI providers for one-click setup.
• Support for custom API parameters, request headers, and disabling model thinking.
• Added context history for improved consistency in continuous subtitle translation.
• Automatic video title detection combined with TMDB information for better translation quality.
• Per-provider daily usage statistics and quota limits (request count and token limits).
• Persistent notification badge in the PiP window when usage quota is reached.

• Screenshot button moved to the settings menu for a cleaner control bar.
• Fit-to-window button now displayed by default.
• Learning Mode guide page redesigned with two-column layout and responsive small-window support.
• DRM-protected videos automatically fall back to standard Picture-in-Picture mode.
• Cross-origin PiP now shows a login prompt for unauthenticated users.

• Fixed black screen when using PiP on certain streaming sites (e.g. friDay Video).
• Fixed cross-origin video prompt overlay not being clickable.
• Fixed subtitles not applying correctly after free user downloads.
• Fixed missing login prompt when backend returns 401 unauthorized.
• Fixed Learning Mode guide page overflowing in small windows.