Show HN: 오픈소스 유튜브 요약, 자막 및 채팅, 타임라인 사이드패널
hackernews
|
|
📦 오픈소스
#ai
#ai 딜
#anthropic
#openai
#오픈소스
#요약
#유튜브
#크롬확장프로그램
원문 출처: hackernews · Genesis Park에서 요약 및 분석
요약
오픈 소스 크롬 확장 프로그램 'QuickSummarize'가 공개되어, 사용자가 유튜브 영상의 자막을 기반으로 AI 요약 및 질의응답을 수행할 수 있게 되었습니다. 이 도구는 크롬 사이드 패널에서 대화형 워크스페이스를 제공하며, OpenAI 및 Anthropic과 호환되는 API를 연동해 영상 내용을 분석하고 자막을 SRT 형식으로 내보낼 수 있습니다. 현재 크롬 웹 스토어 등록은 어려워 개발자 모드를 통한 수동 설치가 권장되며, 사용자는 플랫폼의 자동화 탐지를 피하기 위해 자막을 직접 켜두고 사용하는 것이 안전합니다.
본문
QuickSummarize is an open-source Chrome extension for working with both YouTube videos and ordinary webpages in Chrome Side Panel. It opens in Chrome Side Panel as a source-aware workspace. On YouTube, it uses a transcript-first workflow for summaries, chat, timeline chunks, and subtitle export. On ordinary webpages, it can summarize the current page and answer questions about the current page context through the same side panel. The current release supports YouTube watch pages plus normal HTTP(S) webpages. | Overview | Webpage | |---|---| - Generate AI summaries for YouTube videos - Chat about the current video in a transcript-first workspace - Summarize ordinary webpages in the side panel - Chat about the current webpage or current text selection - Show timeline-based subtitle segments for YouTube - Export subtitles as SRT-formatted text files for YouTube - Support English and Chinese UI - Work with OpenAI-compatible APIs - Work with Anthropic-style APIs - Platform support: YouTube watch pages and ordinary HTTP(S) webpages - Browser support: Chrome / Chromium browsers with Side Panel support - Distribution: source install only for now I do not currently expect this project to be reliably accepted in the Chrome Web Store, so the recommended installation method is loading it in Developer Mode. - Open a YouTube video or an ordinary webpage - For YouTube, manually turn on captions in the player - Open the extension side panel - Use the workspace tabs to summarize, chat, inspect the timeline, or export subtitles when available By default, the extension does not try to open captions for you. There is an optional setting to auto-try opening captions, but it is disabled by default because it may look like automation behavior to YouTube. Automatic caption opening is not recommended because it may require the extension to interact with the YouTube player, trigger extra caption requests, and behave more like automation than a normal user action. That can make subtitle retrieval less stable and may increase the risk of being flagged by platform defenses. For the safest workflow, manually turn on captions first, confirm they are visible on the video, and then use QuickSummarize. git clone https://github.com/EchoTide/QuickSummarize.git cd QuickSummarize npm install npm run build - Open chrome://extensions/ - Turn on Developer mode in the top-right corner - Click Load unpacked - Select the extension folder in this repository If you do not want to build locally, you can download a packaged archive from GitHub Releases. - Open the repository Releases page - Download the latest quicksummarize-vX.Y.Z.zip - Unzip it locally - Open chrome://extensions/ - Turn on Developer mode - Click Load unpacked - Open the unzipped folder and select the inner extension folder that containsmanifest.json The release zip contains an extension/ directory, so Chrome should be pointed to that inner folder after extraction. This repository can publish release packages automatically. When a tag like v0.1.0 is pushed, GitHub Actions will: - Install dependencies - Run tests - Build the extension - Package the extension folder into a zip file - Attach that zip file to a GitHub Release Example: git tag v0.1.0 git push origin v0.1.0 After loading the extension: - Open the extension settings page - Fill in: Provider API Base URL Model API Key Language - Save the configuration Provider notes: OpenAI-compatible : uses{baseUrl}/chat/completions Anthropic-style : uses{baseUrl}/messages with standard Anthropic-style SSE events Speed recommendations: - For OpenAI-compatible endpoints, prefer smaller fast models such as the gpt-nano class or similar lightweight variants - For other vendors, prefer their low-latency flash class models when available - Larger reasoning models can work, but summary/chat streaming will usually feel slower in the side panel Optional: - Enable Automatically try to open captions (risky) only if you understand the risk of automation-like behavior - Enable Selection translation with DeepL if you want the floating translate button after selecting text - Add your own DeepL key and optionally choose a fixed translation target language - For translation, you need a DeepL API plan/key; this feature calls DeepL directly and does not use your LLM provider key When a YouTube video is active, the side panel exposes a transcript-first workspace with three modes: Summary : generate or regenerate a readable summaryChat : ask questions about the current video through a transcript-first agent loopTimeline : inspect transcript chunks and refresh timeline output When a normal webpage is active, the side panel exposes a page workspace with: Summary : summarize the current page or selected textChat : ask questions about the current page in a page-context agent flow - Open a YouTube video page - Turn on captions manually in the YouTube player - Confirm captions are visible on the video - Open QuickSummarize - Click Summarize - Open a YouTube video page -
Genesis Park 편집팀이 AI를 활용하여 작성한 분석입니다. 원문은 출처 링크를 통해 확인할 수 있습니다.
공유