Shorts Lab
上传长视频 → AI 检测最佳 30秒钩子片段 → crop to 9:16 vertical → burn TikTok-style subtitles → download MP4. Everything runs in your browser. Free alternative to OpusClip and Munch.
将视频拖到此处
MP4、MOV、WebM — 浏览器可以播放的任何格式
最佳效果：2–20分钟的面对镜头或访谈内容
Whisper AI 转录
转录让工具检测最佳钩子并生成字幕。 如果不需要字幕，请跳过。
Select the spoken language for accurate subtitles.
First-time download only · cached in browser afterwards
Transcript
↕ Drag on preview to reposition crop
✨ AI-Suggested Hooks
Click to jump to the best 30s windows detected by AI.
Hooks appear after transcription.
🎬 Frame & Crop
Or drag the preview above to fine-tune manually.
💬 Subtitle Style
📝 Edit Subtitles
Click any segment to fix transcription errors.
Transcript appears here after transcription.
导出您的短视频
The free alternative to OpusClip and Munch
Tools like OpusClip and Munch charge $20–150/month to take a long video, find the best moments, reframe to 9:16 and add subtitles. They process everything on their servers with a monthly clip quota. Shorts Lab does the same workflow entirely in your browser, with no monthly limit and no server costs — because your CPU does all the work.
4步工作流程
- Upload — drop any video file. MP4, MOV, WebM — anything your browser can play.
- Transcribe — Whisper AI (same model used by OpenAI's API) transcribes the audio with word-level timestamps. Or skip this step if you just want the crop.
- Edit — the AI suggests the 3 most "hook-worthy" 30-second windows based on speech density. Pick one, or drag the timeline handles to any custom range. Preview with live subtitle rendering.
- Export — WebCodecs encodes a full-quality MP4 with subtitles burned in, ready to upload to TikTok, Instagram Reels or YouTube Shorts.
钩子检测
We use a sliding-window algorithm that scores each 30-second segment by the number of words spoken per second. High speech density correlates with high-energy moments — the parts where the speaker is most engaged, making their point, or delivering value. The top 3 windows are ranked and shown as clickable suggestions.
字幕
Subtitles are rendered directly onto the video frames during export using the Canvas API. Three styles are available: TikTok (bold text on a semi-transparent box, yellow accent — the most viral-looking style), Minimal (white text with a drop shadow, clean and readable), and Solid Box (colorful background, high contrast).
隐私
Your video never leaves your device. All transcription, cropping and encoding happen locally using WebAssembly and WebCodecs. The only network requests are the one-time model downloads (Whisper model from HuggingFace, both cached by your browser after the first run).
限制
- Export requires WebCodecs — Chrome/Edge 94+ or Safari 16.4+.
- Transcription requires a model download (75–145 MB) on first use.
- Encoding speed depends on your CPU — expect 1–3 minutes per 30-second clip.
- Very long videos (>60 min) may run out of browser memory during audio decoding.
常见问题
- 短视频实验室会上传我的视频吗？
- 不会。所有视频处理都使用 FFmpeg.wasm 在您的浏览器中完成。您的视频永远不会离开设备——完全私密和安全。
- 支持哪些短视频平台格式？
- 支持 TikTok、Instagram Reels、YouTube Shorts 等平台的竖屏格式（9:16）。自动将横屏视频转换为最适合移动端观看的竖屏比例。
- 可以调整视频的哪些参数？
- 可以设置输出分辨率、宽高比（9:16、4:5、1:1）、修剪起止时间,以及选择画面裁剪位置（居中、靠左、靠右）。
- 处理后的视频质量如何？
- 使用 H.264 编码输出高质量 MP4 文件,适合直接上传到社交平台。您可以调整输出质量以在文件大小和画质之间取得平衡。
- 短视频实验室免费吗？有使用限制吗？
- 完全免费,没有使用限制、无需注册、不添加水印。处理完全在浏览器中进行,适合内容创作者日常使用。