Subtitles & voice cloning,
on your machine.

Caption. Built for creators, podcasters, and anyone who needs accurate timed text. Editable transcript with draggable caption clips on a waveform; SRT, VTT, plain text, or FCPXML for Final Cut Pro; 16 languages auto-detected, including Mandarin and Cantonese; optional cloud route via OpenAI whisper-1.

Voice Clone. Powered by CosyVoice 3 — runs locally, your reference clips never leave the machine. Zero-shot cloning, no fine-tuning. Nine languages: 中文 · English · 日本語 · 한국어 · Deutsch · Español · Français · Italiano · Русский. Whisper auto-transcribes the reference; transcript is editable inline. Lazy install — model downloads only when you opt in.

• Waveform timeline with per-pixel detail at any zoom (⌘+ / ⌘− / ⌘0)
• Draggable caption clips — body to move, edges to resize
• Inline text edits with auto-grow textareas, ⌘Z undo
• Insert / split / merge / delete — per-row + button (auto-fills the gap), in-row ✂ to split at the caret, ↑ to merge with the row above, ⌘J to join a multi-row range
• Variable playback (1× / 1.25× / 1.5× / 2×) without changing pitch
• Near-instant 简/繁 conversion for Chinese transcripts (OpenCC)
• Plays anything: WebM with Opus + AV1, MKV, FLAC, .ogg via Web Audio fallback
• Local mode never leaves your machine — useful for NDA'd content

The CosyVoice runtime is too heavy to bundle (~6.5 GB), so it installs on demand the first time you download a model. Vuesub detects what's missing and shows install snippets inline. You'll need three tools on your PATH first:

• python 3.10–3.12 — CosyVoice's torch pin. brew install [email protected]
• uv — venv + dep installer, 10× faster than pip. curl -LsSf https://astral.sh/uv/install.sh | sh
• git — shallow-clones the CosyVoice repo. xcode-select --install

Heads up. Voice Clone is fully tested only on Apple Silicon Mac. Windows and Linux binaries are produced from CI but the runtime path hasn't been live-tested on those platforms. Please open an issue if anything breaks.

SRT. SubRip — the universal timed-subtitle format. Plain text, numbered cues, HH:MM:SS,mmm ranges. Players: VLC, IINA, mpv, QuickTime, smart TVs. Upload: YouTube, Vimeo accept SRT directly. Editors: Premiere Pro, DaVinci Resolve, Final Cut Pro, CapCut, Camtasia. Burn-in: ffmpeg -vf subtitles=in.srt.

FCPXML. A real Final Cut Pro project (FCPXML 1.11) where every segment is a native, restyleable <caption> on lane 1 — not burnt-in titles. Frame-aligned to your chosen rate (1080p · 30 / 24 / 25 / 29.97 / 60). Drop-frame timecode for 29.97 and 59.94, NDF for everything else. Import: File → Import → XML… → pick the .fcpxml.

VTT. WebVTT for HTML5 <video> with a <track> element. Same shape as SRT, with . instead of , and a WEBVTT header. Pick this for self-hosted video on a website you control.

Plain text. Just the lines, one per segment. No timing. Use it for the content, not the timing — turning a podcast into a blog post, dropping a transcript into Notion, feeding text to an LLM, sharing notes.