Spaces:
Running
title: README
emoji: 🎬
colorFrom: green
colorTo: yellow
sdk: static
pinned: false
short_description: 'Open-source video pipeline: clipper, AI video & more'
ViralMint
Open-source video pipeline: clipper, AI video & more.
Scout trends, cut clips, write and voice scripts, caption and render, on your own computer.
Website · GitHub · MCP server · YouTube · X
What ViralMint does
ViralMint is a desktop app for macOS, Windows and Linux. It runs a local server and opens in your browser, and the heavy lifting (Whisper, ffmpeg, rendering) happens on your machine.
| Studio | What it does |
|---|---|
| Scout | Find outlier videos in a niche on YouTube, TikTok and Douyin, scored against each channel's median views |
| Clipper | Turn a long video into vertical shorts, with AI-suggested moments, shot-aware face reframing and captions |
| Smart Video | Script → voice → matched stock or AI visuals → word-by-word captions → MP4, editable scene by scene |
| AI Video & Image | Text- and image-to-video, AI images and thumbnails through one account |
| Motion Graphics | AI-authored motion graphics rendered locally with HyperFrames |
| Voice | 400+ text-to-speech voices, and voice cloning that runs fully on-device |
| Tools | 40+ single-purpose tools: captions, subtitles, translate, reframe, remove silence, auto-zoom and more |
| MCP | 150+ tools exposed to Claude Code, Cursor and any MCP client |
Try it in your browser
Word-by-Word Captions is ViralMint's caption engine running in your browser. Drop in a short video and Whisper (through Transformers.js) transcribes it locally; you get TikTok-style captions where each word lights up, a live preview, and .ass / .srt files. Nothing is uploaded.
The clips above show its six presets, left to right with the top row first: brainrot, bold, neon, viral, karaoke and minimal.
A look inside
 Smart Video: a script becomes scenes with a voice, matched footage and captions, editable one scene at a time. |
 AI Video: text- and image-to-video with templates, from budget to flagship models. |
 Motion Graphics: AI-written HyperFrames compositions, previewed and rendered on your machine. |
 Tools: 40+ one-job tools, from captions and translation to reframing and silence removal. |
Open models we build on
- Speech-to-text: Whisper through faster-whisper on the desktop and Transformers.js in the browser
- Voice cloning: VoxCPM, run locally through a ggml port
- Face tracking for reframing: YuNet from OpenCV Zoo
- Motion graphics: HyperFrames by HeyGen
Licence and contact
The open-source core is on GitHub under AGPL-3.0, and so is the captions Space. Issues and pull requests are welcome. Questions: support@viralmint.net