Your first clips, in three steps

1. Give it a video. Paste a link or upload a file on the dashboard. Anything up to 12 hours and 25 GB.
2. Set up the run. Pick a niche, a clip length, how many clips you want, a caption style and an edit style. Every one of these has a sane default, so you can hit go immediately if you want.
3. Export what you like. Clips come back scored and previewable. Export the ones you want, then publish or schedule them.

The whole run happens on our servers. You can close the tab. Come back and the clips are waiting under Your projects.

Picking a source video

The single biggest lever on clip quality is the video you feed in. ClipLab can only find moments that exist.

• Talking beats scenery. Speech is what gets transcribed, scored and captioned. A video with continuous talking gives far more to work with than a montage.
• Longer is usually better. A two hour podcast has more candidate moments than a six minute video, so the top eight clips are genuinely the top eight rather than everything there was.
• Clean audio matters more than clean video. Transcription accuracy drives caption accuracy, and captions carry the clip.
• Paste from anywhere. YouTube, Twitch, Rumble, Kick, X, Vimeo, Reddit, Dailymotion, Odysee, Streamable, or upload the file directly.

Niches and why they change the scoring

A niche is not a tag. It changes what counts as a good moment.

In a podcast, the payoff of a story is the clip. In gaming, the reaction is the clip. In education, a self-contained explanation is the clip, and cutting into the middle of one ruins it. Picking the right niche tells the scorer which of these patterns to look for.

Auto-detect reads the transcript and picks for you, and it is right most of the time. Override it when your video is a hybrid, for example a history channel doing commentary, where you would rather be scored as commentary than as documentary.

Clip length

Auto lets each clip end where the thought ends, which is almost always the right answer. A moment that needs 43 seconds gets 43 seconds.

Use a fixed range when the destination demands it. Under 60 seconds keeps you inside the classic Shorts and Reels window. Ninety seconds to three minutes suits story-driven content where the setup is the point.

Resist the urge to force long clips out of short moments. Padding a 20 second idea to 60 seconds is the most reliable way to kill retention, and retention is what the algorithm actually reads.

Caption styles

Captions are burned in, word-timed and always on. Most people watch muted, so the captions are the first thing that has to land.

• Beasty, heavy and high contrast. Best for fast, loud content.
• Hormozi, bold with key words highlighted. Strong on business and motivational talking heads.
• Modern, clean white. Gets out of the way, good for documentary and education.
• Yellow and Neon, high-visibility accents for busy footage.

Pick one and stay with it. Caption style is one of the few things that makes a channel feel like a channel instead of a folder of unrelated videos.

Edit styles

The edit style drives a motion plan built per clip, from the transcript and the tracking data. It is not a fixed template applied to everything.

• Clean. Captions and a slow push in. Nothing else moves. Use it when the content is serious or when the footage is already busy.
• Dynamic. The default. Cuts on sentence boundaries, punches in on the words that land, progress bar at the bottom.
• Cinematic. Slower cuts, whip transitions, a spotlight on the subject, film grain. Suits storytelling and history.
• Maximum. Everything on, loud and fast. Effective on high-churn feeds, exhausting anywhere else.

Reframing to 9:16

Every clip is tracked, not letterboxed. When there is a face, the crop follows it, and when two people trade lines the crop follows whoever is talking.

When there is no face to track, a vision model looks at the frame, names the subject and boxes it, and the crop uses that box. This is what stops gameplay footage and b-roll from being framed on the middle of nothing. It only runs where the old heuristic would have been guessing, so talking heads never pay for it.

The preview you see uses the same tracked crop as the export. What you preview is what you get.

Trend checks

A clip is only marked as trending when there is measured data behind it, and the badge shows you that data rather than asserting it. Search interest, view velocity on comparable videos, and how saturated the topic already is.

If the check cannot run, because a quota ran out or the topic is too niche to measure, it says so. It does not fall back to a confident guess. A tool that always has an answer is a tool whose answers mean nothing.

Titles, descriptions, tags and the rest

When you publish through ClipLab, the metadata is written from the clip's actual content, not from a template. The description explains what the short is about, because a description that describes nothing is a wasted ranking signal.

Category, audience setting, AI disclosure and playlist are all filled in on upload. You can set an account-wide link block that gets appended to every description, which is where affiliate or channel links belong.

Titles are the one place worth overriding by hand when you have a strong instinct. Everything else is better left automatic and corrected by what the analytics say later.

Scheduling and pacing

Auto-schedule takes every clip you exported and spreads them out. The default is one short every two days, and that default is deliberate.

Uploading fifteen shorts in an afternoon is the loudest spam signal an account can send, and it also wastes them: each one competes with the last for the same initial audience. Spacing costs you nothing and protects the channel.

Once you have posted enough shorts for the data to mean something, ClipLab starts proposing times based on when your channel's own audience actually watches, rather than generic best-time-to-post advice.

How Catalyst learns from your channel

Catalyst is the memory layer. It reads your published shorts' analytics back and turns them into lessons that steer the next run: what to clip, how to frame it, what to title it, when to post it.

• Lessons are weighted by evidence. A short that 60,000 people saw counts far more than one that 200 people saw. There is a hard ceiling on that weight so a single viral video cannot hijack your whole strategy.
• It is allowed to say it does not know. When a video simply was not distributed, Catalyst declines to diagnose it instead of inventing a critique of your pacing.
• Everything is scoped per channel. Lessons from one connected channel never leak into another. A gaming channel and a history channel learn separately, as they should.

This is why the first two weeks on a new channel feel generic and the fourth week does not.

What the price difference actually buys

Every plan has the same toolkit. Trend checking, channel analytics, Catalyst learning, publishing, captions, reframing and edit styles are all there on Starter.

What Pro adds is that none of it waits for you. Pro watches the channels you pick and starts clipping within seconds of a new upload, schedules a whole batch across days at proven times, posts on its own, and folds each result back into what it picks next. On Starter you do those four things yourself, and the tool is just as capable when you do.

Pick Starter if you are hands-on and post in bursts. Pick Pro if you want to stop being the bottleneck.

Turning a long-form channel into a shorts channel

ClipLab treats a channel that was built on long-form differently from one that was always shorts-first, because the goal is different.

On a shorts-first channel, the short is the product. On a repurposed long-form channel, the short is a trailer, and its job is to send people to the full video. The descriptions and end framing change accordingly.

If you have a back catalogue, this is the cheapest growth available to you. A nine hour video that got 300 views is not a failure, it is 40 shorts you have not cut yet.