Turn a long video into shorts automatically
An hour-long podcast or stream has a handful of moments that would do well on TikTok, Reels and Shorts. Finding and cutting them by hand takes an afternoon. Frameloom listens to the whole recording, picks the moments that stand on their own, and turns each into a 9:16 clip with subtitles — after you've chosen which ones to make.
Free to try: 5 reels from a 20-minute 1080p video costs 28 tokens, and you get 30 every week.
How to do it
- Open Auto reels and drop your file in, or pick it from your device. It stays on your device.
- Check the settings. Choose how many reels to look for, how long they should be, and the framing: the whole picture on a blurred background, or cropped to fill the screen. Then pick, retitle and trim the moments it found.
- Press Run, then preview and download the result.
How the moments are picked
Speech recognition runs in your browser and writes down every word with its timing. The transcript — just the text, never the video — goes to a language model, which reads it like an editor would: a strong opening line, a complete idea, a story with a payoff, advice someone can use.
Every suggestion is checked before you see it: clips start and end on whole sentences, fit the length you chose, and never overlap. If the model is unavailable, the pace and shape of the speech are used instead, so you always get suggestions.
You get the list with the words spoken in each moment and why it was picked. Preview any of them on your video, rename them, move where they start or end by a second at a time, and untick the ones you don't want.
Each reel is cut, framed to 9:16, given subtitles timed word by word — the spoken word highlighted, or one word at a time — and the title over the top for the first seconds as a hook. All of it is rendered on your own device.
Questions
- Is my video uploaded?
- No. The video never leaves your device: recognition and rendering both run in your browser. Only the transcript text is sent, to pick the moments, and it isn't kept.
- How long can the video be?
- Up to 60 minutes. Recognition takes a few minutes for a long recording on a computer with a modern graphics card, longer without one.
- Which languages work?
- English, Romanian, Russian, Ukrainian, Spanish, French, German, Italian, Portuguese, Polish, Dutch and Turkish. For languages other than English, choose the Accurate recognition model.
- Can I change what it picked?
- Yes. You see every suggestion before anything is made: rename it, move its start or end, or leave it out.
- What if the speaker isn't in the middle of the frame?
- Choose the blurred background: the whole picture is kept, with the title above and the subtitles below it.