YouTube Create’s:
AI Editing
allows you to select up to:
25 videos or photos at once.
Seeing the number 25,
it’s easy to think:
“I’ll just throw them all in and let the AI pick.”
Today, just change one habit:
Don’t.
Before clicking Next,
first sort your materials into three piles:
Must keep.
Candidates.
Delete.
This step takes only one minute,
but is often more effective than:
regenerating the First Draft repeatedly.
Why shouldn’t you interpret “up to 25” as “best to use all 25”?
YouTube’s official statement is that:
AI Editing can select up to:
25 items.
This is a:
technical limit.
It’s not an official recommendation that:
more clips equal better AI editing.
For example, if your video is a:
20-second Short,
but your phone has:
8 clips of coffee brewing,
6 shots of latte art,
5 of the shop interior,
4 of desserts,
and 2 product photos,
throwing it all in means the AI must guess:
Which latte art is most important?
Which interior shots are duplicates?
What should open the video?
So today’s method isn’t about:
teaching AI to be smarter.
It’s about:
giving it fewer unnecessary options upfront.
First pile: Must Keep
Ask yourself:
“If this video could only have three shots, which three absolutely can’t be missing?”
For example, a:
30-second dessert Short.
Must keep clips might be:
1. Hook
Cutting into cake with filling oozing out.
2. Key action
Pouring hot chocolate sauce.
3. Final result
The finished dessert placed on a table.
If these are missing,
the video’s:
main message
changes completely.
So put them into:
Must Keep.
“Must Keep” isn’t just about the prettiest shots
The real criterion is:
does the story break without this shot?
For example:
A beautiful shop interior
might look great.
But if removing it
still lets viewers fully understand:
what the dessert is,
how it’s made,
and what the final look is,
then it’s probably not:
Must Keep.
Instead, it’s:
Candidate.
This distinction is very important.
Second pile: Candidates
Next, sort clips that:
“make the video better if included but aren’t crucial.”
For example:
A second angle of coffee being poured.
A waiter bringing the dessert to the table.
Close-ups of appearance.
Customer reactions using a fork.
Background environment.
Packaging shots.
These materials help:
fill pacing,
add emotion,
and give context.
They can be put into:
Candidates.
Candidates are best left for AI to choose
These shots usually don’t have:
a single “right” choice.
For example,
left-angle shot of a coffee machine
or
right-angle shot—
both work.
This is where AI editing is most:
helpful,
making a first draft arrangement
when multiple reasonable options exist.
Humans then review:
whether pacing feels smooth.
Third pile: Delete
This pile is fastest to identify.
Remove clips that show:
Complete duplicates
The same action filmed five times
with nearly identical shots.
No need to keep them all.
Obvious failures
Out of focus.
Finger blocking the view.
Suddenly shaky footage.
Subject moving out of frame.
No new info added
If the previous clip already shows the full product,
and the next is merely the same angle showing it again,
with no new action,
result, or emotional change,
move it to:
Delete.
The most important question today
For every clip, ask:
“What new thing does this clip add?”
Is it new:
information?
action?
emotion?
result?
twist?
If yes, it may stay.
If no,
“it’s basically the same as the previous clip.”
Move it out.
Don’t fine-cut at this stage
This step is not about:
trimming every 7.3-second clip precisely down to 2.1 seconds.
Doing so means you’re doing the AI’s job for it.
Today’s goal is just about:
selecting footage.
No need to:
color grade.
add captions.
match to the beat.
or even
finalize order exactly.
Just sort into:
Must Keep,
Candidates,
and Delete.
Example demonstration
Imagine you just filmed a:
30-second smart door lock Short.
You have:
18 clips on your phone.
Must Keep
Holding hands full, standing at the door.
Finger touching the fingerprint lock.
The door unlocking successfully.
These three form:
the complete cause and effect:
problem,
solution,
and result.
Candidates
Walking up to the house door.
Close-up of the bag.
Close-up of the door lock.
Person noticing the lock.
Indicator light turning on.
Looking back after entering.
These can help:
improve pacing,
but not all are mandatory.
Delete
Four close-ups of the same door lock.
One is out of focus.
One has a finger blocking it.
Two are nearly identical.
Keep only the best one.
Move the rest out.
So you may reduce from:
18 clips
to:
8 truly useful clips.
Using 8 clips isn’t an official YouTube recommendation
Just to be clear,
YouTube doesn’t specify:
“best to limit to 8 clips.”
The official max is still:
25 items.
8 clips is just the result in this example workflow.
Some videos need:
only 4 clips.
Some may need:
15 clips.
The true principle isn’t:
fixed numbers,
but:
keep only footage that serves a purpose.
Then hand over to YouTube Create
After sorting three piles,
the clips to feed into AI editing are mainly:
Must Keep + Candidates.
Next:
open YouTube Create.
Select:
Create Shorts.
Add materials.
Choose:
Voiceover,
Music,
Captions,
and more.
Then let the AI:
generate the First Draft.
YouTube positions this feature as:
organizing raw footage
into a
customizable first draft.
This is why pre-sorting materials is so important.
After the AI draft, don’t rush to re-edit everything
When the first version is ready,
just check three things first.
Did the Must Keep clips all stay in?
If a key:
hook
was deleted,
add it back first.
Are there too many candidates?
If the same action appears:
two or three times in a row,
delete one.
Are any Delete clips still included?
If you cleaned well before,
this should be rare.
This is the point of the three-pile method.
It doesn’t guarantee:
AI gets it perfect first try.
But makes the:
first draft much easier to improve.
Why is this more effective than repeatedly regenerating?
Suppose the first draft is bad.
If you don’t adjust the material,
and just regenerate,
the AI just reshuffles the same:
25 options,
which may change the order
but won’t remove problematic items like:
duplicate footage,
weak shots,
and unnecessary clips.
If you tidy up:
the input first,
the AI’s next task is much simpler.
This is similar to text AI.
Longer prompts aren’t always better.
What matters is:
clear and relevant context.
For a 15-second Short, “Must Keep” should be stricter
YouTube Create currently supports:
under 15 seconds,
15 to 30 seconds,
and 31 to 60 seconds.
The shorter the video,
the more essential each clip’s role is.
If only 15 seconds,
you may only have room for:
Hook,
problem,
solution,
and result — four key points.
Background shots,
second close-ups,
or third reaction shots
might not fit.
So before choosing materials formally,
know roughly:
the intended length of your video.
60 seconds doesn’t mean you should keep everything
Longer videos
can hold more content,
but that doesn’t mean:
duplicate shots suddenly gain value.
If the same point can be made in:
10 seconds,
don’t force a:
40-second stretch
just because you have 60 seconds.
Short videos truly need:
each clip to have a reason to keep viewers watching.
The key benefit of this method is “human priority setting” before AI editing
AI can analyze:
images,
sound,
rhythm,
and footage,
but it can’t know:
what you care about most today.
For the same product video,
your main goal might be to:
highlight convenience.
For another,
focus on appearance.
Even with identical footage,
“Must Keeps” might differ.
So the three-pile method is really about:
telling yourself, before AI starts editing,
what is truly essential and must not be lost.
Next year’s Gemini conversational editing will also need this step
YouTube announced on September 23:
a global rollout in early 2027 of:
Conversational Editing Assistant.
In the future, you can tell Gemini:
Trim,
reorder,
adjust pacing,
add text,
and keep chatting to refine.
This is obviously more convenient.
But:
if your raw footage has:
20 almost identical shots,
the conversational AI
still needs to guess which you prefer.
So no matter how powerful new tools get,
prioritizing your materials
is still essential.
The most common mistake: marking too many clips as “Must Keep”
If you organize:
18 clips,
and:
15 are marked as:
Must Keep,
that’s basically:
no sorting at all.
Must Keep should really mean:
clips that change the story if missing.
Not:
“I worked hard filming this, so I don’t want to delete it.”
One of the hardest things for creators is:
cutting their own footage.
AI won’t solve this for you.
A simple starting ratio to try
This isn’t a YouTube rule,
but a practical guideline from SasaDaily.
Suppose you have:
12 source clips.
You can try:
Must Keep
3 clips.
Candidates
5 clips.
Delete
4 clips.
Then feed:
8 clips
to the AI for drafting.
If you feel the story is missing something,
bring back one clip from:
Delete or candidates.
This is usually easier than:
putting all 12 clips in upfront
then cutting down later,
because you learn the purpose of each clip from the start.
The true test: “Draft retention rate”
If you want to know:
whether this method works,
you don’t need complex dashboards.
Just note:
how many clips you end up keeping in the AI draft?
For example:
You submit 8 clips,
and keep 6.
This suggests your material selection was good.
If you submit 20 clips,
but only keep 5,
next time you might want to:
do earlier selection.
This ratio isn’t a quality score,
but a way to see:
if you’re giving AI too many unused clips.
Another metric: how long is human editing?
The real efficiency measure isn’t:
how fast AI makes the draft,
but how long from:
raw footage
to
final video.
If:
AI generates a draft in 10 seconds,
but you spend 35 minutes fixing it,
the overall effect is limited.
If sorting materials first means:
AI drafts,
then you only spend 10 minutes finishing,
that’s a real:
workflow improvement.
Don’t chase “full AI auto editing” just yet
Spend this one minute:
opening your album.
Before uploading to YouTube Create,
sort footage into:
Must Keep: clips essential to keeping the story intact.
Candidates: clips that improve pacing but aren’t mandatory.
Delete: duplicates, failures, or clips adding no new info.
Then only give AI:
footage that really serves a purpose.
AI’s best role isn’t:
deciding “what your video is really about.”
It’s, once you know what’s most important,
helping you:
make the first draft faster.
If you want to know which step in your workflow is best given to AI first, comment "workflow."
Today, let’s improve a bit with AI.
Learn one AI skill daily.
Save a bit of time every day.
Enhance your ability little by little each day.
SasaDaily, growing together with you.
