Clip live television without guessing.
The scrubbing strip under the player is accurate to what is actually on air, so the moment you mark is the moment you get. A rolling week of recording behind it means you can still cut something that aired on Tuesday.
The strip is pinned to real broadcast time, over a week of recording. Mark in, mark out, and the clip is the frames you saw.
Every live stream arrives late, and it arrives late by an amount that changes minute to minute. So the thumbnail strip you are dragging along is not showing you the same instant as the video playing above it. You mark what looks like the start of the answer and you get the last three words of the question. Then you nudge and re-cut and nudge again, and a thirty-second clip takes six minutes.
The second problem is that the moment you needed has usually gone. Live windows are short; by the time the newsdesk asks for the exchange from the top of the hour, the stream no longer holds it. ClipAI solves both by keeping its own recording of everything that came down the line, and by pinning every thumbnail to real broadcast time rather than to an estimate.
13 things it does. Open any one for how it works.
The thumbnails under the player are locked to real broadcast time, so what you see on the strip is exactly the frame you will get. Mark in, mark out, done — no nudging afterwards.
This sounds like a small thing until you have cut fifty clips in an evening. Removing the correction step is most of the speed difference between clipping live and clipping a file, and it is the reason a producer can work at the pace of the programme instead of falling behind it.
Everything a feed carried is held on your side for seven days. The clip requested on Friday can still be cut from what aired on Tuesday morning.
Seven days is the default rather than the ceiling — retention is a setting, and desks that need longer on a particular feed simply set it longer. It works the same way whether the feed reaches you as a satellite hand-off, a contribution link from the field, or a public stream.
Watch a row of live sources side by side and pull a clip from any of them without switching pages. Built for election nights and long breaking stories, when one producer is holding four channels.
Watching is always on; the analysis of any given tile is something you switch on for that tile when it starts to matter. Keeping that deliberate stops a wall of eight feeds turning into a wall of eight streams of notifications.
As a feed runs, points worth reviewing are flagged on the timeline — a line of breaking news, a quotable answer, a sharp exchange, a change of subject. They are pointers, not clips.
A producer watching a two-hour hearing cannot keep the same attention for the whole of it. Flags are there to bring you back to the minutes that mattered while there is still time to use them, not to decide anything on your behalf.
One clip, marked once, delivered as widescreen for the bulletin, vertical for the phone, square and portrait for the feeds — each properly composed rather than cropped from the middle.
Turning a wide two-shot into a vertical clip normally decapitates somebody. This follows whoever is actually talking, stays clear of the lower third, and falls back to a sensible composition when there is nobody to follow.
A fixed crop never works across a whole programme — what matters in the frame changes shot by shot, and so does the reframing.
Clips can be captioned from what was said and carry a second line in another language underneath — useful when the same clip has to serve a domestic bulletin and an international feed.
A clip stores its trim points, format, branding and captions rather than a finished video. Change your mind a week later and it is rebuilt at full quality from the original — nothing is ever a copy of a copy.
It also means a rebrand does not orphan your back catalogue. Change the lower third and every clip that has not been rendered yet simply comes out in the new look, without anybody re-cutting anything.
AI identifies long silences, stumbles and technical breaks. Keep cuts as suggestions for a producer to review, or enable automatic trimming for that step.
A pause before an answer can carry meaning. Keep review enabled for sensitive interviews, and choose automatic trimming where it suits the material and your workflow.
A four-minute exchange can be split into a numbered run of clips that make sense in order, for the platforms where nobody watches four minutes.
A companion app for NewsMind, so a producer watching a feed on the way home can mark a moment and send it back to the desk without opening a laptop.
Shows the stories your newsroom is running today next to whether any clip exists for them yet. A quiet list of what the desk has written about but never cut.
Out of the box a finished clip is uploaded to its destination and held one step short of the audience, waiting for someone to release it. Where a desk wants a route to run unattended, auto-publish can be switched on for that destination.
The two are not all-or-nothing. Most desks run the highlights account automatically and hold anything involving a named person for review, which is the arrangement the per-destination setting exists to support. Turning it on is a permission in its own right, and the record shows for every published item whether a person released it or a rule did.
A satellite hand-off, a contribution link from the field or a public stream — connected once and then always there, recording quietly whether anyone is watching or not.
Sit with the feed live, or open it hours afterwards and scroll back through the day. Both work identically, because you are always working against the same recording.
Two keystrokes while the programme is still running. Trim to the frame afterwards if you want to, but most clips do not need it.
Pick the shapes you need, let it reframe the vertical, add captions, apply the brand. All of it is a set of choices on the clip, so any of it can be changed later.
The clip is delivered to its destination as a draft — private, unpublished, held. It sits there until somebody with the authority approves it.
Your guest says the thing at 20:14 and the programme runs until 20:45. You mark it as it happens, take widescreen and vertical, let it reframe onto whoever is speaking, and stage it to the channel's accounts. By the time the credits roll the clip is sitting in the queue, dressed and captioned, waiting for the output editor to release it.
A rival channel leads on a remark from a press conference you also carried. You open the feed, scroll back to 11:00, find the exchange on the strip and cut it — even though the live window closed hours ago. Fifteen minutes from being told about it to having it staged.
Results, the leader's headquarters, the opposition's headquarters and the national broadcaster, all on one wall. Concessions and claims of victory come in unpredictable order across all four. You clip from whichever tile speaks first, without closing anything or losing your place on the others.
Flagged moments, suggested trims, proposed titles and recommended platforms are all proposals. None of them change a clip until you accept them.
Automatic cutting from flagged moments delivers clips into a review queue rather than to an audience. Whether that queue then feeds a destination automatically is a separate decision you make per destination.
The suggested crop is a starting point you can drag. On a shot where the story is the person not speaking, you will want to.
Cutting a clip and putting it in front of an audience are separate permissions. A junior producer can prepare all evening without ever being able to publish.
Nothing reaches an audience by accident or by default. A clip is held in the least public state its destination allows until a person releases it — or until a rule you deliberately switched on for that destination does.
Pauses are marked for review, never removed. The pause before a difficult answer is frequently the most important thing in the clip.
Exports are disposable and rebuilt on demand. The recording underneath them is what is kept, so quality never degrades through repeated versions.
Where a score or a recommendation is based on general practice rather than your own audience data, it says so on screen instead of dressing a guess up as a measurement.
Out of the box, a finished clip is uploaded and held — private, unpublished, waiting, even on platforms that have no draft state of their own. That is the default because it is what a desk wants in its first month. Once you trust a route, auto-publish can be switched on for that destination alone: the highlights account runs unattended while anything naming a person still waits for an editor. What does not change is the record — every published item shows whether a person released it or a rule did, and switching a rule on is itself a permission somebody had to hold.
Satellite hand-offs, contribution links from the field, and public streams, plus recorded files you upload. Once connected they all behave the same way inside the studio — the same strip, the same rewind, the same clipping.
Seven days is the default because it covers the overwhelming majority of requests, but the window is a setting and can be as long as a feed warrants. Storage is the only real constraint.
The recording and the instructions, both of which are small compared with keeping a rendered copy of every version of every clip. Any clip can be rebuilt exactly as it was, or rebuilt in a new brand, at any time.
No. Many newsrooms use the studio purely for cutting and hand the finished clips to their own social team or their own publishing system. Where platforms are connected, they are connected once by an administrator and reused per account.
The useful demonstration is the one that runs on your footage and your wire feeds. Tell us how your desk works and we will show you the parts that matter — and say so where we do not help.