{"full_name":"Huanshere/VideoLingo","owner":"Huanshere","name":"VideoLingo","shard":"1","url":"https://github.com/Huanshere/VideoLingo","description":"Netflix-level subtitle cutting, translation, alignment, and even dubbing - one-click fully automated AI video subtitle team | Netflix级字幕切割、翻译、对齐、甚至加上配音，一键全自动视频搬运AI字幕组","homepage":"https://docs.videolingo.io","topics":["ai-translation","dubbing","localization","video-translation","voice-cloning"],"tag_pages":["dubbing","localization","voice-cloning"],"language":"Python","license":"Apache-2.0","avatar":"https://avatars.githubusercontent.com/u/147359634?u=712048e3e39786a2d875f72f78467fbfbcfa29e5&v=4","created_at":"2024-08-09T11:19:52Z","pushed_at":"2026-11-13T10:58:15Z","metrics":{"stars":19269,"forks":2127,"watchers":105,"issues":458,"prs":179,"releases":50,"issues_open":18,"issues_closed":440,"prs_open":7,"prs_closed":77,"prs_merged":95,"discussions":0},"growth":{"day":{"stars":18,"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":[19134,19178,19210,19219,19239,19251,19269],"forks":[2112,2119,2122,2123,2125,2126,2127],"watchers":[null,106,106,106,105,105,105],"issues_open":[null,18,18,18,18,18,18],"prs_open":[null,7,7,7,7,7,7]},"boards":{},"same_owner":[],"issues":[{"number":441,"title":"这个项目的一些局限性","comments":4,"created_at":"2025-04-25T15:03:47Z"},{"number":375,"title":"苹果芯片使用videotoolbox硬件加速FFmpeg视频合成","comments":4,"created_at":"2025-01-24T16:30:16Z"},{"number":330,"title":"功能需求：希望能加入源字幕做校对的步骤","comments":2,"created_at":"2024-12-07T13:28:02Z"},{"number":314,"title":"gpt 的  prompt 不合适","comments":2,"created_at":"2024-12-03T06:31:27Z"},{"number":634,"title":"处理字幕的时候速度比V2版慢很多","comments":1,"created_at":"2026-10-14T21:47:58Z"},{"number":547,"title":"可否加入火山引擎的tts","comments":1,"created_at":"2026-04-09T03:30:06Z"},{"number":533,"title":"这个新版v3.0.1的配音效果是不是不如以前了？","comments":1,"created_at":"2026-03-12T14:51:37Z"},{"number":524,"title":"大佬能原生自持indexTTS2.0吗？","comments":1,"created_at":"2026-01-16T01:33:46Z"}],"discussions":[],"latest_release":{"name":"VideoLingo v3.2.0 🎨","tag":"v3.2.0","published_at":"2026-11-13T08:57:18Z","body":"# VideoLingo v3.2.0 🎨\n\nSubtitles that look right in every language, an optional PyTorch-free engine for local Qwen3-ASR, and an installer that leaves your global pip settings alone.\n\n### 🚀 What's new\n\n- **Built-in multilingual subtitle typesetting.** Burned-in subtitles no longer use one system font for everything. Each of 33 languages has its own font, size and spacing, whether it is the translated row or the source row. Lines are broken by their rendered width, measured with HarfBuzz, preferably at punctuation and between words; Chinese, Japanese and Thai are split into words with BudouX. Arabic, Devanagari, Thai and other complex scripts are shaped correctly whatever fonts your system has.\n- **The fonts come with VideoLingo.** 4.4 MB of OFL fonts ship in `fonts/`. The Chinese and Japanese faces are too large for the repository: they are downloaded once, on first use, and checked by SHA-256, with jsDelivr mirrors as a fallback. Docker images fetch them at build time.\n- **Subtitle length is now a width.** `subtitle.max_length` is measured as the line is drawn: a Latin letter counts about 1, and 80 is one full landscape row (about 35 Chinese characters on the source row). The new defaults are `max_length: 80` and `target_multiplier: 1.0`.\n- **The classic filter still works, and no longer runs off the screen.** Set a `font_name` to use your own font with FFmpeg's `subtitles` filter, as before. The same filter is the fallback when the typesetter can't be used, and it now gets long Chinese, Japanese and Thai lines pre-broken so they stay inside the frame.\n- **Optional audio.cpp engine for local Qwen3-ASR.** With `whisper.qwen_engine: audiocpp`, recognition and forced alignment go to an audio.cpp server you run yourself (GGUF weights, no PyTorch); its address is `whisper.audiocpp_url`. It is opt-in: `auto` never picks it.\n- **The installer keeps your pip configuration.** `--auto-mirror` used to run `pip config set global.index-url`, which changed the package index for every Python project on the machine. The chosen mirror is now used for this install only.\n- **See where your models are.** `installer.py --check` and a finished install list the Hugging Face cache in use with the models already in it, the local model folder, and where FFmpeg comes from. The start guide explains how to reuse models you already have, including those of a standalone WhisperX.\n- Fixed \"Translation not found\" warnings for TTS provider names in the sidebar.\n\n### ⚠️ Before upgrading\n\n- **Install the new dependencies** `uharfbuzz`, `budoux` and `fonttools`: run the installer again, or `pip install -r requirements.txt`. Until then, subtitles are burned the old way and a warning asks you to re-run the installer.\n- **The first Chinese or Japanese burn needs a network connection** to download that font (10–20 MB, once). Offline, put the variable font from google/fonts or fonts.google.com in `fonts/`, for example `NotoSansSC[wght].ttf` or `NotoSansJP-VariableFont_wght.ttf`.\n- **If your `config.yaml` still has `max_length: 75` and `target_multiplier: 1.2`**, click **Restore defaults (20/80)** under *Subtitle length tuning* to use the new values.\n- `font_size` now scales the typesetter's own sizes (15 and 17 are its defaults), and the typesetter does not use `margin_v`. Setting a `font_name` turns the typesetter off for every language.\n- If an earlier `--auto-mirror` install changed your global pip index, `pip config list` shows it and `pip config unset global.index-url` removes it.\n\n### ✅ Release validation\n\n- Automated suite: **832 passed, 20 skipped** (macOS arm64, Python 3.12, the FFmpeg installed by the installer).\n- Subtitles were burned with FFmpeg and libass and compared before and after for Chinese, Japanese, Korean, Russian, Arabic, Hindi and Thai, for the typesetter and for the fallback filter.\n\n---\n\n# VideoLingo v3.2.0 🎨\n\n每种语言的字幕都排得好看；本地 Qwen3-ASR 新增可选的免 PyTorch 推理引擎；安装器不再改动全局 pip 设置。\n\n### 🚀 这次更新\n\n- **内置多语言字幕排版。** 烧录字幕不再所有语言共用一种系统字体。33 种语言无论作译文行还是原文行，都有","is_prerelease":false},"readme":true,"readme_zh":false,"news":[],"stale":false,"snapshot_at":"2026-11-16T09:23:14Z","history_start":"2026-09-06"}