An AI video cutdown is an edit plan that keeps one approved idea and removes beats so a 12-second master can become a 6-second version and a 3-second version. It is not a new generation. It is not an aspect-ratio change. You approve one product story, then you decide which moments can leave without turning that story into a different ad.
The master is the long approved cut, planned here at about 12 seconds. The 6-second version and the 3-second version are survivors of that same idea. A survivor can drop the open, shorten the proof, or fold a closing hold into the offer frame. It cannot invent a new promise, a new location, or a new product angle and still count as a cutdown.
Cliprise is the AI creative platform where that master gets generated, from a prompt or from a product still, inside a marketing workflow that already includes image generation, image-to-video, and text-to-video. Generate the approved master with the AI Video Generator. Spend credits on a replacement short only when this plan says a trim would break the idea. Read the credits shown for that run before you launch it. An edit of a file you already approved does not, by itself, ask the model for a new render.
Leave duration, frame shape, and file settings on their own pages
Teams stall when one brief tries to solve four different jobs. This page solves one job: which beats remain when the clock gets shorter. The other three jobs already have pages. Link them. Do not rebuild them here.
How long to generate lives in video duration limits for 5, 10, and 15 second generations. Use that page when you pick a render length. Come back here when the idea is approved and you need a 6-second and a 3-second edit. Twelve, six, and three are edit targets, not a claim that every model exposes those durations. If the model offers a nearby length, build the master there and apply the same removal logic. Empty frames added to force a round number are not beats.
Frame shape is a composition job. A 9:16 placement and a 1:1 placement need their own framing so the product stays in the safe area. That work lives in aspect ratio mastery. Cropping a 16:9 master into a vertical file is a reframe. It is not a cutdown. A cutdown can be vertical and still be a cutdown, but only because you removed time, not because you sliced the sides off.
File settings are a delivery job. Codec, resolution, frame rate, captions, and playback checks live in the AI video export settings checklist. Lock the beat plan before anyone exports a delivery copy. A shorter file that uses the wrong frame rate is still the wrong file, and a correct export of the wrong beats is still the wrong ad.
If a brief says "make it vertical, 3 seconds, and 1080," split the note. Aspect goes to the aspect page, the edit length stays here, and the file spec goes to the export page. Generation duration, when you still need a render, goes to the duration page.
Write the idea sentence before you touch a beat
A cutdown stays honest only when one sentence is true at every length. Write it in plain language, with the product and the action a shopper can see:
A shopper can see this sunscreen bottle and the pump that shows it is ready to use.
That sentence is the law for the 12-second master, the 6-second lift, and the 3-second version. If you cannot say it over the pictures that remain, you do not have a shorter version of the ad. You have a different ad.
The sentence should name a visible fact. "The brand feels premium" is not a sentence you can check on a mute frame. "The label faces camera and the pump completes one press" is a sentence you can check. Keep claims inside what the picture shows. A planning example can show a bottle and a pump. It should not invent clinical results, user counts, or a promise the label does not carry.
Write the sentence once, at the top of the beat sheet, and do not revise it in the edit to rescue a trim. If the trim needs a new sentence, the failure rule has already tripped. A new short still serves the original sentence unless the strategy actually changed. The new generation changes the composition, not the idea.
One idea also means one product. A master that opens on a bottle, proves a spray, and offers a jar is three products. Pick the bottle. Read the sentence aloud over each cut. If the reader has to add "and then you would have seen the bottle," the cut failed.
Beat sheet for a product ad: open, proof, offer
A product ad that will be cut down uses three beats. Open. Proof. Offer. Each beat has a picture job, a time box on the master, and a note about whether it is allowed to die.
The open puts the product in a place the shopper understands. For a bottle, that might be a shelf, a bathroom counter, or a hand entering frame. The open is setup. It earns its time only if the product is already readable before the beat ends. An open that hides the bottle until a later beat is a trap: the 6-second cut will want to delete the open, and the product will vanish with it.
The proof shows the product doing one job. One pump, one pour, one smear of texture, one cap coming off. A second proof is a second ad. Proof is the beat people are most reluctant to cut, and it is also the beat most likely to bury the label. Hands, foam, and motion cover type. Plan the proof so at least one frame still shows the product at a size a stranger can recognize.
The offer holds the product and the line the viewer must remember. On a master, that can be a steady label plus a short end hold. The line might be the product name already printed on the pack. You do not need a new slogan to have an offer beat. You need the pack large enough, and still enough, to read. If a dedicated closer is part of the master, build that hold with the AI video end card handoff. On the 6-second and 3-second versions, a separate card often has no time. Fold the closer into the offer frame instead of adding a fourth beat.
Fill the times before you prompt. A 12-second master can use these boxes:
| Beat | Time on the master | Picture job | Allowed to die? |
|---|---|---|---|
| Open | 0.0 to 2.0 | Place and product, readable by the end of the beat | Yes, if frame one of a later beat is already the product |
| Proof | 2.0 to 8.0 | One use of the product | Only the weak middle. Keep one clear action if the offer still needs it |
| Offer | 8.0 to 12.0 | Product large, line readable, hold | No. It can get shorter. It cannot disappear |
Two plus six plus four is twelve. If your times do not add up, the sheet is a wish, not a plan. Fix the arithmetic before you generate.
A beat may die only when the idea sentence stays true and the product and the offer stay visible without it. If the open is the only readable label, it cannot die. If proof is the only moment that shows the job, it cannot die. The offer can get shorter. It cannot disappear. "Yes" in the last column needs that reason. "Yes" alone is how products get buried.
Inside proof, mark approach, action, and result. Keep the action. Drop the approach. Keep the result only when it shows the product better than the offer. One proof only. If the action covers the label, speeding the master up will not save the short version. A cutdown removes pictures. It does not squeeze every picture into a blur.
Match the end of the open to the start of proof: same side of the bottle, same light, same hand. That match is what lets you delete the open later. The offer frame should already work as a paused, muted still, the same standard as AI video first frame design. Shorten the offer by cutting hold time, not by cutting the product out of frame. Put necessary words on the pack, or in a line that still fits the shortest survivor.
What survives at 6 seconds and at 3 seconds
Treat the two short versions as lifts from the master, not as new scripts. A lift keeps selected ranges. Those ranges do not have to be neighbors on the timeline. They do have to add up to the target, and the join between them has to show the same product in the same idea.
This is the survival table for a product ad built on the boxes above. The sunscreen walkthrough later uses these rules on a specific bottle. The table is the rule. The bottle is one illustration.
| Beat | 12-second master | Survives at 6 seconds | Survives at 3 seconds |
|---|---|---|---|
| Open | 0.0 to 2.0. Place, hand, product readable | Cut, or a half-second arrival only if frame one is already the product | Cut. A 3-second version has no setup budget |
| Proof | 2.0 to 8.0. One action, with approach, action, and result | Lift the clear action, about 3.5 seconds. Drop the approach | Keep only if that slice still shows a readable product and the offer. Otherwise the lift fails |
| Offer | 8.0 to 12.0. Product large, line held | Lift about 2.5 seconds of the same frame. Shorten the hold | Must live inside the surviving frames. A later card does not fit |
| Separate end card | Optional, and only if the offer still has the product | Fold into the offer hold | Leave it out |
| New story | Not allowed | Not allowed | Not allowed |
Check the sums. A 6-second plan of 3.5 seconds of proof plus 2.5 seconds of offer is 6.0. A 3-second plan has to name its own 3.0 seconds. "Use the best bits" is not a sum.
Delete the open first when proof already starts on the product. A typical 6-second lift keeps 2.0 to 5.5 of the action and 8.0 to 10.5 of the same offer frame. The join has to read as a move closer on the same bottle. A new room or a surprise top-down pack is a new scene, and that length has failed.
Three seconds can hold one picture with a small move, or two pictures that both already contain the product. A legal lift starts with the label readable and keeps the action beside the type. If the label is covered, stop. A head trim drops the start and keeps the rest in order. A tail trim drops the end, which deletes the offer when the name lives there. A lift keeps ranges that are not neighbors. A new generation is what you do when none of those pass. The bottle example is that case at 3 seconds.
Generate a master that can lose beats
Generation comes after the sheet, not before. The prompt, or the still you animate, should describe beats you are willing to delete. If you ask a model for "a cinematic morning, a beach, a smile, a logo," you will get a film with nothing removable, because nothing carries the product.
Cliprise can help you generate that master as text-to-video when you are still finding the motion, or as image-to-video when a product still must stay recognizable. Use the Image-to-Video AI Generator when the label, the bottle shape, or the pack shot is already approved as a still. Animating a known frame keeps the offer honest. Asking a fresh text prompt to invent the same label is how type melts and how the short versions lose the only readable frame.
Models differ in clip length, how closely they follow a still, and how steadily they hold small type. Check the duration controls and the credits shown on the run you actually launch. If one pass cannot cover the master, generate open, proof, and offer as separate approved shots that share the idea, the product, and the light, then assemble them in your editor. The cutdown plan still applies. Each shot is not a new story.
Design the first frame of proof and of the offer as if they might open the short versions. A master can recover from a soft opening. A 6-second lift that starts on proof cannot. Mark in and out points on the file you received. If the pump lands at 3.4 seconds instead of 2.0, move the sheet. The survivor sums still have to land on 6.0 and 3.0.
Record the run in an AI video generation log: filename, prompt or still source, model route, the duration you requested, aspect, and the idea sentence. Add two fields this workflow needs: the 6-second survivor ranges, and the 3-second decision (lift or new generation). A later test is comparable only if those fields stay filled.
Illustrative walkthrough: a sunscreen bottle
The scene in this section is illustrative. It is a planning example for a plain pump bottle of mineral sunscreen, not a campaign report, and it cites no prices or performance results. The useful action is one pump. The offer is a readable label, not a medical claim. The idea sentence is:
A shopper can see this sunscreen bottle and the pump that shows it is ready to use.
The 12-second master
Open, 0.0 to 2.0. The bottle stands on a bright bathroom shelf. A hand enters from the right. By 1.5 seconds the front label faces camera and is large enough to recognize as a sunscreen bottle. The background is a simple tile, not a lifestyle story. Light stays even so a later close-up will match.
Proof, 2.0 to 8.0. The hand takes the bottle and presses the pump once. Product lands on the back of the other hand. You can see the pump finish. You cannot read the label during the press, because the fingers wrap the front of the bottle. That occlusion is deliberate in this example. It is the defect that will matter at 3 seconds. From 2.0 to 3.2 the hand approaches. From 3.2 to 5.6 the pump presses and product appears. From 5.6 to 8.0 the hand settles and the bottle drifts slightly away from camera. The settle looks pleasant on a 12-second master and has no job in a short cut.
Offer, 8.0 to 12.0. The bottle returns large, label forward, pump clear of the type, and holds. The last 1.5 seconds are that same frame, not a second slogan card. A dedicated end card, if the placement needs one, is a master-only piece from the end-card handoff, and it is already marked disposable for the short versions. The master passes the idea sentence. Approval does not mean every shorter lift will work.
The 6-second lift
Delete the open. Lift proof from 3.2 to 6.7 (3.5 seconds: the press and the first product on skin) and the offer from 8.0 to 10.5 (2.5 seconds of the hero label). The sum is 6.0. The join goes from the bottle in hand to the same bottle, larger, label forward, in the same light. The idea sentence still fits. Send those points to the editor.
The 3-second attempt that fails
A hopeful lift is proof from 3.2 to 5.2, then offer from 8.0 to 9.0. The sum is 3.0 seconds. The arithmetic passes and the ad does not.
From 3.2 to 5.2 the fingers cover the label. Someone who never saw the shelf cannot read the bottle, and the one-second offer then flashes type they were not prepared for. You cannot say the idea sentence over these three seconds without borrowing the master, and the master will not be playing in that placement. Cropping tighter on the fingers removes more of the bottle. Speeding the whole master keeps every beat and makes each one unreadable. A beach film, a talking creator, and a pack shot would be three new stories. The fix is a new short whose first frame already satisfies the sentence.
The replacement 3-second generation
Generate one new clip of about 3 seconds. The first frame is the hero: label forward, pump clear of the type, bottle large, same bathroom light as the master so the set of files still looks like one campaign. A small motion is enough. The pump presses while the label stays readable, because the hand comes from the side and below, not across the type. There is no shelf, no second angle, and no end card.
The idea sentence does not change. The composition does. Log it as a new generation, child of the master, variable "3-second composition," with a note that the lift failed because the pump occluded the label. The 6-second file is an edit. The 3-second file is a new render of the same idea. Hide that distinction and the next person will treat three pictures as one clean length test.
Use image-to-video from the approved hero still when the label must stay put. Use text-to-video only if you have no still and you will reject takes where the type wanders. Either route runs through the AI Video Generator. Check the duration options and the credits on screen. Keep the prompt narrow: one bottle, one pump, label readable for the whole clip, no scene change. Pause the new short on frame one and read the idea sentence. If the label fails, it fails immediately.
The rule: when a cutdown fails, generate a new short
A cutdown fails when no trim, and no lift, can keep all three of these true:
- The idea sentence is still true if the viewer never saw the deleted beats.
- The product is recognizable in the frames that survive, at a size that does not depend on a shot you removed.
- The offer is understandable on mute inside those same frames, not in a beat that got cut.
Fail any one check and you stop cutting. You generate a new short for that length. You do not "cut harder." Harder cuts are how labels vanish and how offers shrink into a flash of color.
The sunscreen 6-second lift passes. The 3-second lift fails the product check and the offer check. Write "new short" on that row.
The same fail shows up in other sheets. The only readable label sits in the open you wanted to delete. Proof cuts to a logo card with no product, and the short version would have to live on that card. The join changes room, wardrobe, or product color. The short version only makes sense with a voice line. The in and out points cannot sum to the target unless you keep a beat the checks say must die. In each case, stop trimming.
A failed cutdown is a scoped generation. The new short inherits the idea sentence, the product, the light, and the offer, and it changes only the composition the length requires. A new hook, a new audience, or a new claim belongs in the AI ad creative testing workflow, with one variable at a time. Put the rule in the brief: if the lift breaks the idea, hides the bottle, or strands the offer, request a new short from the hero frame.
Three mistakes that look like efficiency
These three moves feel fast. Each one produces files that cannot be compared and often cannot be understood. The sunscreen sheet exists so you can see the difference between a real cutdown and a shortcut.
Cropping the master
Cropping changes the rectangle. A center crop of a horizontal master into a vertical file will often slice the bottle, the pump, or both. The timeline is the same length as before. You have not made a 6-second ad. You have made a narrower 12-second ad that may no longer show the product.
People crop because a placement asks for 9:16 and the master was 16:9, and the deadline treats those as one task. Split them. Build the vertical frame in the aspect-ratio workflow, with the product placed for that frame. Plan the 6-second and 3-second survivors in that same frame. Scaling the hero until the label fills the screen is also not a cutdown. Scale is not time. If the label must be larger at 3 seconds, say so on the sheet as a reframe or a new short. Stretching 16:9 into a square removes nothing and clips type. The export checklist makes separate delivery files. This page chooses the pictures inside each length.
Regenerating three stories
The second mistake is one prompt for a 12-second lifestyle film, a second prompt for a 6-second testimonial, and a third prompt for a 3-second pack shot, launched together because "we need all the sizes." You now have three ideas. If the pack shot performs differently from the lifestyle film, you learned nothing about length. You learned that you shipped three concepts and named them after durations.
Generate the master first. Lift the 6-second version if the checks pass. Generate the 3-second version only as the same idea with a new composition, the way the bottle example does. If you want to test a creator-style proof against a product-only proof, label that test in the creative testing workflow. Give each concept its own idea sentence. Do not hide concepts inside a length label.
The generation log makes the mistake visible. A row that changes prompt, still, model, duration, and story together is not a length test. Cutdown rows share one idea sentence. The 6-second row says "edit" and lists ranges. The 3-second row says "edit" or "new generation," and if new, the variable is composition for time.
Three unrelated films also spend three generations and three review cycles. A cutdown spends the edit. A replacement short spends one more generation. Read the credits shown before you launch it. A softer version of the same mistake is rerolling the master until its opening already looks like a 3-second ad. Let the master stay a master. Rerolling the long clip to solve the short placement throws away beats you already approved.
Burying the product
The third mistake is a master where the product is a guest. The open is a sunset. The proof is a smile. The offer is a wordmark on a colored card. Somewhere in the middle, a bottle sits on a table for under two seconds, small, in profile. On the 12-second timeline you can still claim the product "appears." On the 6-second lift you will keep the smile, because it is the prettiest beat. On the 3-second version you will keep the wordmark, because it looks like an offer. The bottle is gone. The ad is now a mood with a logo.
Look at the 3-second column and ask where the product is. If the answer is "in a beat we cut," the sheet is already a burial. Redesign the offer so the product is the offer, or start a new short on the product. Mood can support a master. It cannot replace the product in the frames that survive.
Burying also happens when the product is present and unreadable, which is the sunscreen pump covering the label. Presence is not readability. Pause the survivor on the first frame, the middle frame, and the last frame. In at least two of those stills the product should be recognizable. If only a blurry middle frame works, fix the frame.
Review the plan, log the decision, then export
A cutdown is ready for the editor when the sheet has numbers, not adjectives. Hand over this packet:
- The idea sentence.
- The master filename and the aspect you actually approved.
- The 6-second ranges, with the sum written out.
- The 3-second ranges, or the words "new generation" plus the still or prompt to use.
- The join note, including what must not be cropped.
- A line that says the end card, if any, stays on the master and comes off the short versions.
- A line that says file settings wait for the export checklist.
The editor assembles the lifts. Abrupt and clear is acceptable. Smooth and product-free is not. If the join needs a new shot, send the failure back. A zoom, a flash, or a stock transition hides a failed check.
Review in order. Read the idea sentence over the master, then over the 6-second lift, then over the 3-second file. If the reader has to add plot, mark the failure before anyone talks about taste. A dull join that passes can be improved inside the same frames. A lively join that drops the bottle does not pass.
Log before export. An edit row points at the master and the ranges. A new-short row has its own filename, the parent master, the still or prompt, the model route, the credits shown, and why the lift failed. A later hook test is a new variable. It should sit beside the cutdown rows, not overwrite them.
Export after that review. The checklist covers the master and each delivery length as separate outputs: aspect you already planned, product safe area, captions that still fit the short hold, and playback of the file itself. If you test the files, length is the only difference between the master and the 6-second edit. Say in the test notes that the 3-second file changed composition. Do not score it as a pure length result.
What you do next
Put the idea sentence at the top of a blank sheet. Block open, proof, and offer so they add up to the master you can actually generate. Mark the 6-second lift. Try the 3-second lift. If it fails the three checks, write "new short" and describe the first frame instead of forcing the trim.
Generate the master, and generate the replacement only when the rule says so, with the AI Video Generator. When the product still is already approved, animate it with the Image-to-Video AI Generator so the label you planned is the label you cut down from. Cliprise is useful here because the marketing workflow stays in one place: the still, the master, the notes in your log, and a new short that spends credits only when a cutdown cannot carry the idea. You can run that work at www.cliprise.app.
Then stop planning and hand the ranges to the edit. A cutdown that stays on the sheet is a wish. A cutdown with in and out points, a passed or failed check, and a log row is a file you can ship beside the master without pretending it is a new story.