How shortshort turns one video into many.
What happens between the drop zone and the MP4, what the AI decides, and what stays in your hands.
What it does
You drop one long video — a talk, a lesson, a podcast. shortshort transcribes it word by word, reads the transcript for complete, quotable sentences, and proposes 1 to 10 distinct shorts of 15 to 90 seconds each. Complete sentences come first: a short can run 25% shorter or longer than your target so it never cuts a thought in half.
Gemini receives the original video with its audio, then selects passages for each requested short. Finishing a sentence takes priority: a ±25% margin is allowed, or 22.5 to 37.5 seconds for a 30-second target. Each short has its own title and edit; selected passages are never reused between shorts.
Every proposal is an edit you can change: cuts, framing, zooms, title and captions. Nothing is rendered until you ask for it.
Import and limits
The Gemini model is selected in the admin settings. If GPT-6 Astra is active, sources larger than 100 MB or longer than ten minutes automatically use Gemini 3.8 Flash when it is available; Astra can analyze smaller sources from images and a transcript. The temporary Google file is deleted after analysis, and Google also applies its own 48-hour expiration.
A lower bitrate or resolution reduces file size without changing the duration. The AI analysis needs an account; manual editing and MP4 export work without one.
Framing and captions
Each clip is cropped to 9:16 around the speaker. Face and scene tracking move the crop when the subject moves; slides and wide shots are detected and kept as a full image. Emphasis zooms stay between ×1.0 and ×1.6, and every keyframe can be adjusted or locked by hand.
For captions, shortshort extracts audio from the selected passages plus a few seconds of context to refine the ending, then sends it to Whisper when an OpenAI key is configured. If synchronization fails, the Gemini timestamps are marked as estimated.
- Impact — bold uppercase captions with one highlighted word per line.
- Karaoke — words light up as they are spoken.
- Minimal — quiet, readable captions without effects. You can also import an SRT file timed to the source.
Export, save, publish
Shorts render in your browser as MP4 (H.264, 1080 × 1920, 30 fps) — use a recent version of Chrome and keep the tab open. Saved shorts from the same video share one source file, so they stay grouped in your library.
Everything is private by default. Publishing shares the short, its original video and the transcript on a public page you can withdraw at any time.