Shorts Lab
Déposez une longue vidéo → l'IA détecte le meilleur hook de 30 secondes → crop to 9:16 vertical → burn TikTok-style subtitles → download MP4. Everything runs in your browser. Free alternative to OpusClip and Munch.
Déposez votre vidéo ici
MP4, MOV, WebM — tout format que votre navigateur peut lire
Meilleurs résultats : 2–20 minutes de contenu face caméra ou interview
Transcription Whisper IA
La transcription permet de détecter le meilleur hook et de générer des sous-titres. Ignorez si vous n'avez pas besoin de sous-titres.
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.
Exporter votre Short
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.
Le flux de travail en 4 étapes
- 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.
Détection du hook
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.
Sous-titres
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).
Confidentialité
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).
Limitations
- 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.
Questions Fréquentes
- Qu'est-ce que Shorts Lab ?
- Un outil pour convertir des vidéos paysage en format vertical 9:16 pour YouTube Shorts, TikTok et Instagram Reels.
- Ajoute-t-il des filigranes à mes vidéos ?
- Non, la sortie est propre sans filigrane ni branding.
- Puis-je prévisualiser le recadrage avant d'exporter ?
- Oui, voyez un aperçu en direct de l'apparence de votre vidéo en format vertical.
- Les vidéos sont-elles traitées sur un serveur ?
- Non, tout s'exécute localement dans votre navigateur pour une confidentialité maximale.
- Shorts Lab est-il gratuit ?
- Oui, conversions illimitées sans frais ni inscription.