Taoye250 downloadsWatch YouTube and local videos with bilingual subtitles, timestamped screenshots, transcript notes, and AI conversations.
Watch, understand and take notes without leaving Obsidian. Open YouTube or local videos, read bilingual captions, capture moments and ask AI about what you are watching.
English · 简体中文 · 日本語 · 한국어 · Español · Français · Deutsch
Desktop Obsidian 1.13.7 or later is required. Mobile is not supported.
Choose a feature below to expand its short GIF and instructions. Each demo covers the named feature.
Watch and organize · Read and translate captions · Ask AI · Keep notes
The recordings use English instructions and highlighted clicks. Cmd/Ctrl means Command on macOS and Ctrl on Windows/Linux. Demo notes and requirements.

Your videos stay in their original locations. The same file is added only once per playlist; files with the same name in different folders stay separate. Rescan updates all folder sources while preserving manually added videos and progress; each Locate folder button reconnects only its own source folder. Mark as unwatched resets selected lessons. You can also reorder videos within a chapter and keep a note for each video.
Duration scanning uses local ffprobe, configurable under Duration settings. Unknown durations can be filled in during playback. Playlist data is stored separately from AI settings in video-library.json. No AI account is needed for playlists or basic playback.





Open a video in the desktop Web viewer core plugin and enable captions on the website. Run Translate video subtitles from Web Viewer, then click Translate transcript with AI. The video stays on the original page; the adjacent panel shows only the toolbar and transcript, with synchronized highlighting, playback controls and clickable timestamps.

Reads accessible HTML5 text tracks; YouTube can use the existing yt-dlp fallback. Reconnect after switching videos or to refresh dynamic captions. Cross-origin players, proprietary or burned-in subtitles and generic web audio transcription are not supported. Web video frame capture is unavailable; AI chat can use the extracted captions.
The Web Viewer demo records the released plugin in Obsidian using original sample material and actual translation results. It is silent; waiting time is shortened.







Open this plugin’s settings and choose AI backend. For OpenAI or Anthropic, enter your own API key. API usage is billed by that provider. For local Codex or Claude Code, install and sign in to that application first; set Codex command or Claude command if it is not found automatically. Local CLI use can still send your content to its AI provider.
Start with Automatic · Economy for supported lightweight models. It caches selections for one hour and never automatically upgrades to a flagship model. Use Manual to choose a specific model; custom API base URLs require Manual mode. If using an API, run Test API connection.
Existing captions and basic playback do not require an AI key. Embedded local captions and sampled frames require ffmpeg/ffprobe; YouTube extraction may need yt-dlp. Install these tools and set ffmpeg command / yt-dlp command if detection fails. Place a matching .srt or .vtt file beside your local video to use sidecar subtitles. Media support depends on Obsidian’s video decoder; try an H.264 MP4 if playback fails.
If a video has no captions, configure No-caption transcription and its separate audio-service credentials; a Codex/Claude subscription alone does not supply that transcription API. Audio transcription can incur extra charges. AI requests send the relevant captions, conversation and any selected frames to your chosen provider. YouTube playback connects to YouTube.
Choose Video screenshot folder and Video transcript folder for exported material. Screenshots are inserted into an open note when the corresponding option is enabled. Local timestamp links need the original file to remain at the same path.
API keys are saved in this plugin’s settings in your vault. Keep those settings private, including when sharing or syncing your vault. If AI fails, check the selected backend, login/API key and model; quota or network errors do not cause an automatic model switch.
For selected text and document translation, use the separate AI Translation Assistant.
Report a problem on GitHub Issues; include the plugin version and error message, but never an API key or private notes.
Local video files you choose may be outside your vault. Local AI integrations also read the CLI’s model catalog and use its existing login in your user directory. The plugin does not install these tools for you.
MIT License · © 2026 Taoye