{"full_name":"FluentRead/FluentRead","owner":"FluentRead","name":"FluentRead","shard":"1","url":"https://github.com/FluentRead/FluentRead","description":"An open-source browser extension for bilingual translation. 一款开源的浏览器双语翻译插件。","homepage":"https://read.thinkstu.com/","topics":[],"tag_pages":[],"language":"TypeScript","license":"GPL-3.0","avatar":"https://avatars.githubusercontent.com/u/317633511?v=4","created_at":"2023-12-22T02:59:51Z","pushed_at":"2026-11-15T15:30:56Z","metrics":{"stars":10931,"forks":624,"watchers":23,"issues":328,"prs":1024,"releases":8,"issues_open":33,"issues_closed":295,"prs_open":10,"prs_closed":34,"prs_merged":980,"discussions":0},"growth":{"day":{"stars":62,"forks":1,"base_day":"2026-11-15"}},"history":{"dates":["2026-11-10","2026-11-11","2026-11-12","2026-11-13","2026-11-14","2026-11-15","2026-11-16"],"stars":[10071,10410,10461,10471,10625,10869,10931],"forks":[561,593,599,599,609,623,624],"watchers":[null,23,23,23,23,23,23],"issues_open":[null,31,31,28,32,36,33],"prs_open":[null,14,16,14,16,15,10]},"boards":{"weekly_rate":140},"same_owner":[],"issues":[{"number":969,"title":"[BUG] 译文嵌入页面部分不稳定","comments":5,"created_at":"2026-10-13T05:57:35Z"},{"number":1341,"title":"YouTube的字幕有时原英文字幕显示两遍，重复","comments":3,"created_at":"2026-11-14T09:04:03Z"},{"number":62,"title":"能不能支持一下Thunderbird（雷鸟）啊","comments":3,"created_at":"2025-05-12T08:40:35Z"},{"number":1344,"title":"【功能请求】视频字幕增加单独显示原文字幕的配置","comments":2,"created_at":"2026-11-14T12:59:15Z"},{"number":1316,"title":"[FR]添加快捷切换翻译服务的功能","comments":2,"created_at":"2026-11-11T10:21:30Z"},{"number":1246,"title":"Hunyuan-MT-7B on SiliconFlow: emoji short texts fail with a glossary error when the glossary is off","comments":2,"created_at":"2026-11-05T11:11:13Z"},{"number":1068,"title":"图片/漫画翻译几乎不可用","comments":2,"created_at":"2026-10-24T05:48:21Z"},{"number":209,"title":"使用nei API无法翻译","comments":2,"created_at":"2026-04-12T06:00:16Z"}],"discussions":[],"latest_release":{"name":"FluentRead 0.0.42","tag":"v0.0.42","published_at":"2026-11-15T14:07:08Z","body":"<!-- fluentread-release {\"schemaVersion\":1,\"workflow\":\"automatic-release-packages\",\"repository\":\"FluentRead/FluentRead\",\"tag\":\"v0.0.42\",\"sourceCommit\":\"6c7754a016890c3be13e583026ef47d69c95c836\",\"manifestSha256\":\"686e65211148609ce279874e0abf0152645276f71baf6d1911b9ab8ddac9f432\"} -->\n\n## English\n\nFluentRead 0.0.42 adds optional local privacy checks for text translation, expands video subtitles and sentence learning, and improves image translation, scanned PDFs, and everyday reading controls.\n\n### Webpage translation and controls\n\n- Choose one-click toolbar translation to translate or restore the current page directly. This is opt-in; the quick panel remains the default toolbar action.\n- In Chrome and Edge, choose whether the translation sidebar belongs to the current tab or the whole window. The current-tab setting is the default.\n- Improved multi-paragraph Bilibili and AI translation fallback handling. Emoji and spacing at paragraph boundaries are preserved locally.\n- Preserve GitHub user and repository names and native page controls. Selection translation stays available when the selection crosses decorative icons.\n- Fixed input-translation restoration and cleanup when pages or panels reopen, so callbacks from closed interfaces do not affect the new view.\n\n### Images and documents\n\n- Choose CPU (WASM) for local PaddleOCR and LaMa processing, or keep automatic backend selection. Runtime details show the processing stage, initialized backend, and automatic CPU fallback.\n- Improved image and manga text recognition, word spacing, wrapping, and placement. Background repair better preserves flat backgrounds, ruled lines, and nearby illustration details.\n- After OCR, scanned PDF text can be selected and copied. Correct recognized source text or re-recognize the current page; uncertain numbers remain available for review against the original.\n- PDF original-layout translations keep their line breaks and placement while zooming. Searchable exports preserve format characters and joined emoji sequences.\n- Fixed batch-import progress, output selection, and ZIP controls in the document sidebar, including narrow-window layouts.\n\n### Video subtitles and learning\n\n- Added native caption timelines for Bilibili, Udemy (including Business), and edX, and support for embedded YouTube players.\n- Float the full-sentence learning card out of a small player, drag it, adjust its height, and dock it back while retaining the current draft.\n- Supported desktop browsers can choose the same local folder to reuse downloaded Whisper models. Existing complete models can be migrated to that folder.\n- Improved handoff when YouTube caption tracks arrive early, and protected legitimate repeated captions from being removed.\n- Bilingual SRT export uses a single timeline and handles overlapping cues.\n\n### Privacy and translation services\n\n- Added translation privacy protection under Advanced options, off by default. Choose local masking with restoration after translation, or block requests containing detected sensitive information.\n- Configure detection categories and custom sensitive phrases. Rule-based detection can miss content; reading-assistant, AI-writing, audio-recognition, and cloud-OCR requests are outside this switch. See the [scope and controls](https://github.com/FluentRead/FluentRead/blob/v0.0.42/docs/en/config/privacy-protection.md).\n- Image uploads through the translation interface are blocked while privacy protection is enabled; use local OCR followed by text translation instead.\n- Added Mistral-hosted GLM and DeepSeek options to the model catalog. Unconfigured AI platforms are hidden by default, and restoring removed services from the custom-service dialog works again.\n\n### Platform compatibility\n\n- Improved Zotero's Windows dropdown keyboard controls and dismissal on outside input, Tab, or window focus loss.\n- Google Drive's native authorization is offered only in confirmed desktop Google Chrome. Unsupported browsers receive clear guidance to use We","is_prerelease":false},"readme":true,"readme_zh":false,"news":[{"title":"FluentRead 上游修复整合与 PR 447 交付验证：从回归缺陷定位到 3937 项确定性流水线验收","url":"https://blog.csdn.net/gitblog_00990/article/details/162487084","source":"blog.csdn","lang":"zh","date":"2026-09-26"},{"title":"开源翻译插件FluentRead获8000+ Star，已接入DeepSeek内核","url":"https://www.80aj.com/2026/09/13/fluentread-deepseek-plugin/","source":"80aj","lang":"zh","date":"2026-09-13"},{"title":"FluentRead智能翻译：3个核心场景+5个进阶技巧解决你的外文阅读难题","url":"https://blog.csdn.net/gitblog_00196/article/details/141485368","source":"blog.csdn","lang":"zh","date":"2026-06-25"},{"title":"FluentRead：英文网页别再复制去翻译站，直接双语对照阅读，6.9k stars","url":"https://www.bilibili.com/video/BV1vkLE6nEEc/","source":"bilibili","lang":"zh","date":"2026-05-16"}],"stale":false,"snapshot_at":"2026-11-16T09:23:14Z","history_start":"2026-09-06"}