Skip to content

v3.0.0 🔥焕新出发,多引擎支持

Choose a tag to compare

View all tags
@github-actions github-actions released this 23 Jun 06:18
v3.0.0
ab25530

v3.0.0 — 全新多引擎与界面重构 / A ground-up multi-engine & UI overhaul

3.0 是自 2.x 以来最大的一次升级：语音识别从单一引擎扩展为多引擎体系，GPU 加速覆盖更多显卡，界面与交互完全重做，任务、校对、合成、翻译四大模块均有质的提升。
SmartSub 3.0 is the biggest upgrade since 2.x — speech recognition grows from a single engine into a multi-engine system, GPU acceleration covers more cards, and the UI/UX is fully redesigned, with major gains across tasks, proofreading, merging, and translation.

当前版本 faster-whisper 未支持 GPU, 如果是 N 卡用户请在高级设置里面，计算设备选择 CPU
The current version of faster-whisper does not support GPU. If you are an NVIDIA user, please select CPU for your computing device in the advanced Settings

🎙️ 多语音识别引擎 / Multiple ASR engines

  • 新增 faster-whisper、FunASR（SenseVoice / Paraformer）、Qwen3-ASR、FireRedASR-AED 等引擎，可按场景与速度自由选择（#313 #314 #316）
    Added faster-whisper, FunASR (SenseVoice / Paraformer), Qwen3-ASR and FireRedASR-AED engines, selectable by scenario and speed (#313 #314 #316)
  • FunASR / Qwen3-ASR / FireRedASR 使用内置原生运行库（sherpa-onnx），无需额外下载运行时；faster-whisper 首次使用时下载独立运行时，模型均按需下载
    FunASR / Qwen3-ASR / FireRedASR run on a bundled native runtime (sherpa-onnx) with no extra runtime download; faster-whisper downloads its own runtime on first use, and models are fetched on demand
  • 每个引擎独立的模型目录，支持「从文件夹导入模型」，并共享 VAD
    Per-engine model directories with "import model from folder", plus a shared VAD

⚡ GPU 加速全面升级 / GPU acceleration overhaul

  • 跨厂商 GPU 加速：CUDA（NVIDIA）、Vulkan（AMD / Intel 等，Windows / Linux）、CoreML（Mac）
    Cross-vendor GPU acceleration: CUDA (NVIDIA), Vulkan (AMD / Intel, on Windows / Linux), and CoreML (Mac)
  • GPU 模式改为「自动」，异步检测修复 Windows 首次启动卡顿；兼容新版 NVIDIA 驱动，修复部分显卡识别不到的问题（#315 #317）
    GPU mode is now "auto" with async detection (fixes the Windows first-launch freeze); compatible with newer NVIDIA drivers, fixing undetected cards (#315 #317)

🎨 全新界面与上手引导 / Redesigned UI & onboarding

  • 全新视觉设计、可折叠侧边栏、命令面板（快捷键唤起）、深色模式与本地化应用菜单
    A new visual design, collapsible sidebar, command palette (hotkey-invoked), dark mode and a localized app menu
  • 重整导航：「引擎与模型」页统一管理转写引擎运行时、语音模型与 GPU 加速，「翻译服务」页集中配置服务商
    Reorganized navigation: an "Engines & Models" page manages engine runtimes, models and GPU acceleration; a "Translation Services" page centralizes provider config
  • 首次启动引导 + 内置示例任务一键试用 + 应用内 FAQ
    First-launch onboarding with a one-click built-in sample task, plus an in-app FAQ

📋 任务管理重做 / Task management redesigned

  • 任务即「可命名工程」，跨重启自动恢复、中断续跑、完成后桌面通知
    Tasks become "named projects" with restart recovery, resume-after-interruption and completion notifications
  • 逐工程的队列与真实的取消 / 暂停；列表 / 网格视图，带视频缩略图、文件大小与预计耗时
    Per-project queues with real cancel/pause; list & grid views with thumbnails, file size and ETA

✍️ 字幕校对增强 / Proofreading editor

  • 虚拟化长列表编辑、键盘快捷键、撤销 / 重做、时间轴行内编辑、搜索 / 替换与失败重译
    Virtualized list editing, keyboard shortcuts, undo/redo, inline timing edits, search/replace and retranslate-failed
  • 覆盖写入前自动保留 .bak 备份，离开未保存有提示；纯转写模式也可使用 AI 校对
    Auto .bak backup before overwrite, unsaved-changes guard; AI proofreading also works in transcription-only mode

🎬 字幕合成到视频 / Burn & mux subtitles

  • 所见即所得的实时字幕预览，跟随播放进度
    Real-time WYSIWYG subtitle preview that follows playback
  • 支持「烧录硬字幕」与「封装软字幕（MKV）」两种输出，可取消并清理半成品；新增导出画质选项（#331）
    Hardcode burn and soft-subtitle MKV muxing, cancellable with partial-output cleanup; added an export quality option (#331)
  • 可选「中文字幕去标点」（源 / 译文），并自动规整简繁字形（#330 #332）
    Optional "remove Chinese punctuation" (source / translation) and automatic Simplified/Traditional normalization (#330 #332)

🌐 翻译能力 / Translation

  • 自定义 OpenAI 兼容服务商配置更稳健、校验更清晰（#326 #328）；服务商支持分组、搜索与测试结果保留
    More robust custom OpenAI-compatible provider setup with clearer validation (#326 #328); provider grouping, search and persisted test results
  • 翻译支持并发批处理，提升长字幕的整体速度（#305）
    Concurrent batch translation for faster long-subtitle throughput (#305)

📥 下载与网络 / Downloads & network

  • 模型 / 运行库下载支持多镜像源（GitHub / gh-proxy 国内加速 / GitCode / ModelScope）与国内优先回退，支持断点续传与校验
    Multi-mirror downloads (GitHub / gh-proxy / GitCode / ModelScope) with domestic-first fallback, resumable and checksum-verified
  • 新增全局网络代理设置（含常用客户端快捷预设与连通性测试）
    Added global network proxy settings (quick presets for common clients + a connectivity test)

🖥️ 平台与其它 / Platform & more

  • ASR 前自动提取视频内嵌软字幕
    Auto-extract embedded soft subtitles before ASR
  • macOS 关闭窗口可后台常驻并在 Dock 显示进度
    On macOS, closing the window keeps the app running with Dock progress

下载 / Download

homebrew (推荐)

安装

brew tap buxuku/tap && brew install --cask smartsub

更新

brew update && brew upgrade --cask smartsub

手动安装

Mac 提示应用已损坏，在终端执行

sudo xattr -dr com.apple.quarantine /Applications/SmartSub.app

夸克网盘 https://pan.quark.cn/s/0b16479b40ca

系统 System 芯片 Chip 格式 Format 下载链接 Download link
Windows x64 exe windows-x64
Mac Apple dmg mac-arm64
Mac Intel dmg mac-x64
Linux x64 deb linux-x64-deb
Linux x64 AppImage linux-x64-appimage
Assets 12
Loading