An AI clipper for people who own the long video

What is an AI clipper?

An AI clipper takes one long video and returns short vertical ones, without a human choosing where to cut. The category exists because the arithmetic of short-form publishing is brutal: the feeds that carry discovery — TikTok, Instagram Reels, YouTube Shorts — want several posts a week, and a single long-form recording contains the material for all of them but in the wrong shape.

Underneath, a real clipper is four decisions stacked on each other, and each one can be done well or badly:

• Where to cut. Which sixty seconds of a ninety-minute recording stand on their own, and where exactly does that passage start and end.
• What to keep in frame. A 16:9 frame recomposed to 9:16 loses about three quarters of its width, so something has to decide which quarter mattered.
• How to caption it. Most short-form is watched muted. Burned-in captions are not an accessibility extra, they are the audio track.
• How to open. The first second decides whether the rest is watched at all, which is what a hook headline is for.

faster-whisper transcribes the source with a timestamp on every word, PySceneDetect finds the shot boundaries, and Google Gemini reads the timestamped transcript together with those boundaries to pick the strongest self-contained moments of 15 to 60 seconds. Each moment is cut with FFmpeg, reframed to 9:16 around whoever is speaking with MediaPipe face tracking, and captioned word by word. The full stage-by-stage version is on how Clippen works.

What separates a good AI clipper from a bad one?

Not the moment detection, mostly. Every serious tool in this category calls a frontier model with a timestamped transcript, and they broadly agree about which sixty seconds of an episode are interesting. The differences show up elsewhere.

Whether it understands the shot. Most clippers assume the subject is a face near the middle of the frame. That is true of a talking-head vlog and false of a podcast with two guests, a webinar with slides, and a stream with gameplay — which is most of the long video that exists. Clippen picks a layout per scene: track one speaker, stack two, stack a shared screen over the presenter, or rebuild a stream's webcam inset as screen-on-top and camera-below.

Whether the cut lands on a word. Word-level timestamps let a clip start on the first consonant rather than a fraction of a second into it. It is a small thing that is instantly audible when it is wrong.

What the unit of billing is. Per minute of source means you know the cost of an episode before you run it. Per finished clip means the cost moves with an output you cannot predict.

Whether it does the whole job. Some well-regarded tools in this category do only part of it. Submagic does not find moments for you. You upload a clip you have already cut and it styles the captions. Going from a 60-minute podcast to finished shorts means running a clipper in front of it, which is two tools and two bills.

Competitor detail checked 2026-07-27. Vendors change their products without notice; verify before buying.

What Clippen deliberately does not do

Two things that competing tools advertise are absent here on purpose. Stating them plainly is a better filter than a vague feature list, because if either is the reason you are shopping, you should buy something else.

No clip "uniquification"

Some clippers advertise that they alter each export so platforms cannot match it against the original. That is duplicate-detection evasion, it is the behaviour platforms remove channels for, and Clippen does not do it. Nobody can promise you will not be actioned, and we will not pretend otherwise.

No bulk auto-posting farms

You can publish clips to TikTok, Reels and Shorts from the dashboard one at a time. There is no queue that machine-guns dozens of accounts, because that pattern is exactly what platform enforcement looks for.

Who is it for?

People clipping their own material. That is the whole design centre, and it explains most of the decisions above.

• Podcasters with a video feed and an archive of episodes that were published once and never cut down.
• Streamers with multi-hour Twitch or YouTube VODs, where the good part is four minutes somewhere in the middle.
• Creators of talks, webinars and courses, where the slides are the content and cropping to the presenter throws it away.
• Interviewers, where the exchange between two people in one shot is the thing worth publishing.

The common thread is that the recording already exists and was already paid for. An archive of long recordings is unpublished inventory: it has reached the audience that already subscribes and nobody else. Clips are how that material reaches the feeds where new audiences, sponsors and subscriptions are found. What any individual clip does there is not something a tool can promise. What is certain is that material which exists only as a two-hour file reaches nobody who is not already looking for it.

What does an AI clipper cost?

Clippen is a paid tool and bills by the length of what you feed it: one credit is one minute of source video, and that one charge covers the whole run — every clip the video yields, the vertical reframing, the captions and the hook. A 60-minute podcast costs 60 credits whether it comes back with six clips or twelve. Plans start at $9.99 a month for 50 credits ($4.99 a month billed annually) and run to $199.99 a month for 1200 credits. There is no free tier.

Concretely: a 40-minute interview is 40 credits, a 90-minute webinar is 90, a four-hour stream VOD is 240. The invoice matches the length of what you fed in, so a month of publishing can be priced before any of it is run.

Why short form at all

• Short-form video delivers the highest ROI of any content format. HubSpot, State of Marketing 2025.
• 91% of businesses use video as a marketing tool. Wyzowl, Video Marketing Statistics 2025.
• YouTube's inauthentic content policy, effective July 2025, names mass-produced and recycled clips as ineligible for monetisation. YouTube Help, Inauthentic content policy.

Those are third-party figures about the format, not claims about Clippen's own results. We publish no user counts, no view numbers and no case studies, because we have none that would survive being checked.