How to Turn a Long Video into Vertical Clips, Locally

What "long video to shorts" actually involves

A two-hour stream, match, podcast, or lecture might contain five clips worth posting. Getting them out takes:

• Selection. Finding those five moments. Done by hand this is the slow part: scrubbing the timeline for an hour to remember where things happened.
• Reframing. Shorts, Reels, and TikTok are 9:16. Horizontal footage has to be cropped, and the crop has to follow the action or the subject wanders out of frame.
• Packaging. Each clip needs to be exported at 1080×1920, usually with captions, since most short-form viewing starts muted.

Step 1: find the moments

Three honest options, in increasing order of automation:

1. Manual scrubbing. Free and precise, but slow. If you go this route, a player with keyboard-driven seeking, speed control, and in/out points makes it far less painful; that is the review workflow FastPlay is built for.
2. Chat or note timestamps. If the source was a stream, chat spikes and your own markers are a decent map of where the energy was.
3. Automated analysis. Tools score the footage and propose candidates. FastClip does this on-device with acoustic and structural signals: speech density, audio energy spikes, dead air, and hook openings. It shows a ranked list of candidate clips with time ranges, and you approve or adjust each one. Automation gets you to review faster; it does not replace your judgment about what is actually worth posting.

Step 2: reframe for 9:16

• Center-crop is the workhorse. It suits centered subjects: a speaker, a streamer cam, most single-subject action. FastClip exports a deterministic center-crop to 1080×1920, so the same clip renders the same way every time.
• Off-center action needs a different framing decision. Subject-tracking crops exist in cloud clippers and full editors; FastClip deliberately does not guess. If the action lives at the edge of frame, adjust the clip range to a moment where it is centered, or finish that one clip in an editor.
• Check the edges. Scoreboards, HUDs, and slide content often live in the 16:9 margins that a 9:16 crop removes. Preview every clip before exporting; if the context is in the margins, that moment may simply not work vertical.

Step 3: captions and export

Most short-form viewing starts with the sound off, so burned-in captions are close to mandatory for talk-heavy clips. The export target for every major platform is the same: 1080×1920 MP4, H.264 with AAC audio. FastClip transcribes speech locally with whisper.cpp and burns captions in at export; the details of that caption workflow are covered in adding auto-generated captions without uploading.

Local tools vs upload-based clippers

Most well-known clipping products (Opus Clip, Vizard, and similar) are cloud services: you upload the footage, their servers process it, and you pay monthly or in credits. That model has real costs:

• Upload time. A two-hour 1080p recording is many gigabytes; on a typical home upload connection, the upload alone can take longer than local analysis.
• Privacy. Unreleased, client, or personal footage leaves your control.
• Recurring pricing. Subscriptions and credit meters, priced for the vendor's GPU bill.

Cloud clippers earn their keep with features local tools do not have yet, like AI virality scoring tuned on platform data and team workflows. But if the job is "get the good moments out of my footage as vertical clips," a local tool does it without the upload, the subscription, or the privacy trade.

The FastClip workflow

1. Import a long local video (it is referenced in place, never copied or modified).
2. Pick the workflow mode that matches the footage, like ActionSports or podcast-style.
3. Let analysis run on-device; FastClip proposes ranked candidate clips with time ranges.
4. Review each candidate, adjust ranges, and select the keepers.
5. Export selected clips as 1080×1920 MP4s, with optional burned-in captions.

Limitations

• FastClip is not an editor. There is no multi-track timeline, no compositing, no B-roll insertion. If a clip needs editing beyond trim, captions, and reframe, finish it in an editor.
• Highlight scoring is signal-based, not clairvoyant. It ranks moments by on-device signals; it does not know your audience. Review is part of the workflow by design.
• Open beta. Expect rough edges; builds ship on GitHub Releases and the planned model is a free tier plus a one-time Pro license for bulk export.