How Snipri works

Snipri turns a long recording where people talk, such as a podcast, an interview, a webinar or a talk, into short vertical clips with captions. You add the video in the Studio and choose how it is clipped. The clips land in My clips, where you can change them and download them.

Once you press Get clips, the job runs on its own, and the Studio shows each step as it starts:

1. Transcribing. The audio becomes a transcript with a timing for every word.
2. Selecting highlights. A language model reads the transcript and scores each moment that could work on its own. Only the moments that clear the bar become clips, and each one keeps its scores and the reason it was picked.
3. Rendering. Each moment is cut from the video, cropped to vertical so the person speaking stays in the frame, and captioned word by word.
4. Done. The clips are in My clips.

With One clip chosen instead of Best moments, nothing is selected: the part of the video you keep becomes a single clip.

Every clip comes back vertical, 1080 × 1920 pixels at 9:16, ready for TikTok, Reels and Shorts.

A job returns up to 10 clips and at most 10 minutes of clips in total, so longer clips come back fewer.

Clips run 15 to 60 seconds by default. Before a job you can choose 10 to 30 seconds, 30 to 60 seconds, 60 to 90 seconds or 90 seconds to 3 minutes instead.

How a moment earns its place is in See why a clip was chosen, and the AI clip maker page shows the selection at work. How the crop follows the people on screen is on the auto reframe page.

Everything after you sign in happens on four screens, each named in the app's menu:

• Studio is where you add a video, choose how it is clipped and follow the job.
• My clips holds every clip, grouped by the video it came from, with the reason each one was chosen, the editor and the downloads.
• Brand keeps the look new clips are made in: a caption style, your logo, a line of text such as your handle, your colours and your own font.
• Plan & videos shows your plan, how many videos you have left and when they come back.

Snipri picks moments by reading what is said, so it needs a recording with speech in it. Music and ambient footage give it nothing to read, and a file without an audio track can't be clipped at all. It needs the picture too, because each clip is framed around the people on screen, so audio on its own is not enough.

Snipri transcribes and captions 25 languages: Croatian, Czech, Danish, Dutch, English, Finnish, French, German, Greek, Hungarian, Indonesian, Italian, Japanese, Korean, Norwegian, Polish, Portuguese, Romanian, Russian, Slovak, Spanish, Swedish, Turkish, Ukrainian and Vietnamese. Choose the spoken language when you add a video, or let Snipri detect it.

• It does not post anywhere. You download the clips and post them yourself.
• It does not predict views or reach. The scores beside each clip say why the moment was kept, not how far it will travel.
• It does not translate. Captions are in the language spoken in the video.
• It has no timeline for joining several clips into one video. The editor works on one clip at a time.

The next guide, Make your first clips, walks through a first job from the start.