How to plan clips from a brand brief with AI
We found 11,748 moments worth cutting in 1,386 source videos of 266 live campaigns, a median 29 seconds long, and flagged 5% of them as a brand risk at 07:25 UTC on 13 Nov 2026.
Read in three passes
Give the model the brief before the footage. A plan built from moments first picks the loudest scenes and checks the rules too late.
- Rules out first. Mark everything the brief bans: topics, words, other brands, footage from elsewhere, other apps' watermarks, AI visuals. One of these sinks a clip however good it is.
- Then what every post must carry: the mentions and tags, a disclosure and where it goes, a link in bio, a sound. These go in the caption, not the cut, but a plan that forgets them fails at the end.
- Then length and format: the shortest and longest clip it pays for, vertical or not, captions or not, and the apps it counts. Now the moments can be judged against all three.
The moments we find
We transcribe each campaign's source videos and score every stretch that could stand alone as a clip, 0 to 100. A moment with profanity, violence or another brand risk is flagged, never hidden.
Fit 9:16 is the share of checked moments whose subject stays in frame at 70 or more of 100 when cropped to vertical.
- Moments
- 11,748
- Score 70+
- 42%
- Fit 9:16
- 34%
- Brand risk
- 5%
One campaign, worked
Honey Hunters of the Himalayas, Doc Clips on Content Rewards. Its page lists 8 rules a post must meet, and its footage holds these moments at the top of our score.
| Moment | Cut | Score | Fit 9:16 |
|---|---|---|---|
| Every Step Is a Calculated Gamble | 10:14 to 10:48 | 91 | 45 |
| Each step a calculated gamble | 10:35 to 11:09 | 89 | 45 |
| Hunters Dangle On Ropes For Honey | 8:55 to 9:07 | 87 | 55 |
- Tag @realmadhoney on TikTok, Instagram, YouTube and X
- Add #madhoney, #honeyhunters and #nepal on TikTok, Instagram, YouTube and X
- Keep captions relevant, natural, and curiosity-driven
- Write "mad honey" in the caption
- Write "Small amounts only. Not for everyone." in the caption
- Put realmadhoney.com/clip in your bio
- Use the words "mad honey"
- Keep clips 10 to 45 seconds
Hold each moment to those rules before it goes in the plan. A low fit means the subject drifts out of a centre crop, so check the frame or pick the next moment. The hooks and captions guide takes the plan from here.
Why clips get rejected
| Rule | Briefs | What to do |
|---|---|---|
| Reposting others' clips | 36% | Cut your own from the source, never a clip already posted. |
| Bought views or bots | 35% | Post from a real account and let the views come. |
| Low effort reuploads | 24% | Add a hook, captions and a tight cut. |
| Misleading claims | 24% | Keep the speaker's words in context. |
| Footage from outside | 24% | Cut only from the footage the brief names. |
| Showing the brand badly | 22% | Skip moments that mock the brand or its people. |
| Off topic content | 20% | Keep every clip on the campaign's subject. |
The planning prompt
You get a clipping campaign's brief and a list of moments from its footage with their times. First list the brief's rules in three groups: what the clip must include, what it must avoid, and its length and format. Then pick the 5 moments that fit every rule, best first. For each, give the cut range, why it works in one line, the rule it is closest to breaking and a hook idea. Leave out any moment that breaks a rule, and say which rule.Paste the brief and a moment list after it. With ClippingAlpha connected, plan_clips does both reads for you: it returns the brief's rules, the moments best first with their cut ranges, and the steps for each clip. Ask the model to hold its picks to the rules, then cut them on our server or by hand.
Common questions
What should I read first in a clipping brief?
3 things before any moment: what the brief rules out, the required caption and tags, and the length and format. A great moment that breaks one of them is rejected.
How long should a clip from a campaign be?
The moments we found run a median 29 seconds at 07:25 UTC on 13 Nov 2026. 217 of 586 briefs we read set a length, and theirs comes first.
Why do clips get rejected?
For breaking the brief. The rules live clipping briefs state most often are reposting others' clips (36%), bought views or bots (35%) and low effort reuploads (24%), across 586 briefs we read at 07:25 UTC on 13 Nov 2026.
Can AI pick the best moments for me?
It can rank them. Of 11,748 moments in live campaigns' footage, 42% score 70 or more of 100, and plan_clips lists each with its cut range and 9:16 fit.
Should I trust an AI clip plan as is?
No. Watch each range before you cut it, check the speaker is not cut mid sentence, and hold every clip to the brief's 3 core rules: content, caption and format.