Video Companion

WHAT'S NEW IN 0.3 - Redesigned side panel with a playback-synced “Now” card, cleaner filters, and more compact navigation - Change output language directly from the side panel: 12 languages or the video's own language - Choose whether to re-analyze the current video or save a language change for future videos—Companion never spends tokens silently - Improved transcript freshness across YouTube navigation and the updated 2026 transcript panels - Improved provider setup: the selected provider and API key are now shown first - Accessibility improvements for keyboard navigation and icon controls Companion is an AI side panel for YouTube videos you actually want to watch. It reads the video's captions and creates timestamped cards for unfamiliar people, places, products, technical terms, and references. Each card gives you a short explanation and lets you jump directly to the relevant moment. As the video plays, Companion highlights the card for the current moment and keeps it in view. Instead of replacing the video with a wall of text, it gives you context while you watch. PERSONALIZED TO WHAT YOU KNOW A doctor, software engineer, and English learner may need different explanations from the same video. Choose from five profiles: - General audience: explains anything non-obvious - Designer / Developer: skips basic web and design terms - Learning English: also flags idioms, slang, and cultural references - AI/ML expert: skips familiar AI terms and surfaces less familiar tools and references - Custom: describe your own background and interests Use “Know it” when a card is too basic and “More like this” when it is useful. After enough feedback, Companion can offer to distill those preferences into a sharper profile. It always asks before making that additional AI call. HOW IT WORKS 1. Open a YouTube video with captions. 2. Open Companion and click “Analyze this video.” 3. Companion sends the video's title and transcript to the AI provider you selected. 4. Browse the timestamped cards, follow them with playback, read the summary, or ask questions about the video. Analysis is manual by default. Optional Auto-analyze can be enabled in Settings. Cached videos reopen instantly without another provider call. FEATURES - Playback-synced cards with a Follow toggle - Clickable timestamps that jump to the matching moment - TL;DR, readable summary, and timestamped video arc - Ask tab with transcript-grounded answers and clickable timestamps - Local Library of previously analyzed videos - Markdown export for individual analyses - JSON backup and restore for the full Library - 12 output-language choices, plus “Same as the video” - Output-language control available directly in the side panel - Optional Auto-analyze, disabled by default - Token usage, estimated cost, provider, model, and cache status shown in the interface - Light and dark appearance support USE YOUR OWN AI PROVIDER Companion supports Claude, OpenAI, Google Gemini, and DeepSeek. You supply the API key for the provider you choose. Keys are stored locally in Chrome and are sent only to that provider when making an API request. Provider pricing and availability vary by model, transcript length, and account. Companion shows usage and an estimated cost in the side panel. Long-video requests require confirmation before the call is made. PRIVACY Companion has no advertising or analytics service. API keys, preferences, feedback, and cached analyses are stored locally in your browser. The current video's title and transcript are sent only to the AI provider you selected, either when you click Analyze or when you explicitly enable Auto-analyze. Companion does not send that content to a Companion-owned server. WORKS BEST FOR - Lectures, podcasts, technical talks, and documentaries - Foreign-language videos when you want explanations in another language - Videos outside your area of expertise - Content where you would normally pause to look up unfamiliar terms Captions are required. YouTube may not expose transcripts for some private, age-restricted, or otherwise unavailable videos.