Which of the 4 AIs Should I Ask? | Comparing Idea Generation for 5-Second AI Videos

Which of the 4 AIs should I ask?

In my previous article I entered the same idea generation prompt into four different AIs:

• ChatGPT

• Gemini

• Copilot

• Claude

I tried them all.

Even when asking the same thing, 'Think of ideas for a 5-second video,' there were significant differences in the ideas produced and the reasons for recommending them.

The reason I'm making 5-second videos is... I'm participating in this project!
It's a project to generate a 5-second video and set it as a thank-you GIF for likes.
Participants will receive 300 credits for Pollo AI.

So, what happened when I threw the same prompt at the four AIs?
To give you the conclusion first, here is how I recommend choosing based on this comparison.

If you're wondering, 'Which one should I ask in the end?', I think it's easy to start by choosing one based on these criteria.

ChatGPT | For those who want to balance personal style and feasibility

ChatGPT doesn't just list interesting ideas; it considers:

• Why that idea suits me

• Whether it works within the short 5-second timeframe

• Whether it is unlikely to fail with video generation AI

• Whether it is easy to use for my future content

It even thought through and proposed that for me.

For example, in my article from two posts ago, GPT's recommended plan combined the travel and Ishigaki Island ocean content I've shared so far with the gratitude for the 'likes' I receive from readers on note.

In other words, it didn't just use the idea of 'it's summer, so let's do the ocean,' but rather...

'What Rin🦁 usually shares' x 'Gratitude for the likes I want to convey this time'

It connected my own content with the video concept in that way.

Furthermore,

It even included advice on points to watch out for when actually making a 5-second video.

It was distinctive not just for the ideas themselves, but for the fact that it also thought together about 'whether that idea can really be made as an AI video'.

Therefore,

'I want an idea that feels like me. But I want an idea that can be properly completed as a video.'

I think the ChatGPT plan is easy to use for people who feel this way.

By the way, GPT recommended the following three plans.

1st Place: A blue heart carried by the waves
2nd Place: Putting readers' likes into a treasure chest
3rd Place: A postcard from a trip

I already did the 1st place 'blue heart carried by the waves' 🩵 in my article from two posts ago, so... this time I tried the 2nd place one,
putting readers' likes into a treasure chest.

Pixverse V5, image-to-video, used 5 credits.
Even if there's a little unnaturalness in the direction,
with the constraint of only using the starting and ending images,
Pixverse V5 is the one that keeps my character the most intact...!

Gemini | For those who want to prioritize visual appeal and impact

Gemini suggests,

• A trunk opening and a character jumping out

• Swimming underwater as a mermaid

• Appearing from inside a sunflower with a 'peek-a-boo!'

and similar ideas where the change in movement is clear at a glance were selected.

This is the interesting part.

For example,

If it's just 'a character standing in a sunflower field,'

it might be a beautiful image, but there is no major change.

On the other hand,

if they 'suddenly appear from inside a sunflower with a peek-a-boo!'

then,

'nothing there'
↓
'a character pops out'

a clear change is created within those 5 seconds.

The idea of a trunk opening and a character jumping out is the same; it is a structure where the 'beginning' and the 'highlight' are easy to understand even in a short time.

Gemini itself also clearly indicated the following as reasons for selection:

• whether it is memorable

• whether it fits easily into 5 seconds

• whether it is easy to reproduce with AI video

.

Judging by the responses this time, the impression is that creating an easy-to-understand highlight within 5 seconds is prioritized over explaining a story in detail.

Therefore,

'I want to make a video that is fun the moment you see it'
'I want ideas that are dynamic and flashy'

For those people, the Gemini proposal seems to be the right fit.

However...
having my 'original character' pop out from nothing is difficult for a beginner to do with low credits.
It didn't go well when popping out of a closed trunk either.

The change between the beginning and the end is too large, so it's not suitable for video generation with low credits... > <
Well, if you don't mind Rai-chan getting a little distorted, it was possible but...
that's not right, is it? I thought...
After all, for this event, my own style was also one of the criteria right?

It seems like it would be possible if you used high credits and created a large number of storyboards with 'Image to Video', but
I think it's impossible to do it with low credits using 'Frame to Video' with just two images for the start and end.

Can you tell...
Both the sunflower field and the trunk,
my videos do not start from 'nothing'...

With the motto of 'It can be done even with low credits!', I'm just glossing over it...
Yes, this sunflower field is also
Pixverse V5, Frame to Video, 5 credits!

Copilot | For those who want easy-to-understand ideas in a short time

Among the four, Copilot's explanation was relatively short, and I understood the main points immediately.

The recommendations this time were,

• Picking up a glowing seashell on a summer beach

• Handing over a train ticket at a summer station

• Summer scenery sparkling inside a kaleidoscope

as proposals.

I think it's easy to imagine the finished video for these three just by reading the text.

The elements are very simple.

Without having to read through multiple stages of explanation,

'Oh, it's a 5-second video like this'

you understand it immediately.

It did not provide answers that explain production precautions in detail like ChatGPT, or delve deeply into the meaning of the work like Claude.

However, because of that, there is less reading burden, and it is faster to look at the candidates and make a decision yourself than others.

Therefore,

“I want to see the ideas quickly first”
“I want to choose what interests me immediately rather than reading long explanations”

I think the Copilot suggestions are easy to use for people who feel this way.

I use Copilot for the ideas, but... I consult ChatGPT about whether they are feasible. I also had ChatGPT create the video prompts. Surprisingly, this is also Pixverse V5, frame-to-video, 5 credits.

Claude | For those who value the worldview and the meaning of the story

Claude recommended ideas such as:

• “Stamping a ‘Like’ on a passport”

• “Unfolding a map to reveal summer travel destinations”

• “Drawing a ‘Like’ with the starry sky and firefly light”

It recommended these ideas.

What was characteristic here was that it didn't choose them simply because they were “pretty” or “would look good in a video.”

For example,

“Stamping a ‘Like’ on a passport”

is one such idea.

If you look only at the movement, it is a very simple act of “stamping.”

However, Claude gave it the meaning of connecting travel content with gratitude for the ‘Likes’ received from readers.

(Pixverse V5, image-to-video, 8 credits. The 3-credit increase is due to the 720p resolution)

Also,

the idea of "unfolding a map to reveal summer travel destinations"

was not just a video of a map changing, but a proposal that makes you feel "the journeys you've taken so far and the world that will unfold from here."

Throughout its response, it also carefully articulated

• how travel content and gratitude connect,

• the reason why only that person can create it,

• and what kind of lingering impression remains after watching.

In other words, I felt that Claude places significant importance on

Claude,

not just "what moves," but "what kind of meaning to give to that footage."

I felt it places significant importance on that.

Therefore,

if you want to create a fantastical and beautiful video,
or if you want ideas that connect to your own activities and creative world,

I think Claude's suggestions are suitable for you.

If you're lost, this way of choosing is fine.

You don't need to try everything.

Just choose one according to what you want right now.

• I want ideas that are less likely to fail
  → ChatGPT

• I want flashy and fun visuals
  → Gemini

• I want to see candidates as quickly as possible
  → Copilot

• I want to create a beautiful world or story
  → Claude

However, there is one thing to note here.

This does not mean that

“ChatGPT always prioritizes feasibility”
“Claude is always fantastical”

is a universal rule.

This is simply thetrend observed when inputting the same prompt into four AIs and comparing their responses.

Even with the same AI, responses change depending on the model used, the information provided, and the history of the conversation.

Even so,

“which AI to consult first”

is a useful guideline to have.

If you have time, it is also interesting to ask multiple AIs

You can generate ideas with just one AI.

However, if you have the time, I also recommend inputting the same prompt into two or more AIs.

In this comparison,

• ChatGPTconsiders feasibility

• Geminicreates visual highlights

• Copilotsummarizes ideas concisely

• Claude is forgiving meaning to the world-building

—that is the difference I observed.

For example,

If you see an "interesting motion idea" from Gemini and think,

"I love this!"

you can take that idea to ChatGPT and

"Please create a structure for this content that won't easily break in a 5-second AI video."

you can consult it like that.

Conversely, after thinking of a framework that is easy to create with ChatGPT, you can also

"How can I make the direction more fantastical and lingering?"

ask Claude that as well.

Thinking about it this way, there is no need to have one AI decide the correct answer from start to finish.

Use each one as a participant in a planning meeting with different areas of expertise.

You can use them in that way too.

And finally,

you choose the idea that makes you think that.

AI will expand your ideas as much as you want.

However,

Which idea should I adopt?
What should I keep, and what should I discard?
What do I want to create?

It is humans who decide that.

I think that including that 'selection process' is what makes creating with AI so interesting.

I won't be entering this time either.
Since I received additional credits, I plan to try out the high-end models at a later date.

Summary of past articles in the series

Why don't you try making a 5-second video too?

☕ Afterword

The one where my character doesn't fall apart the most was
Pixverse V5, image-to-video, 5 credits
.

Even with Pollo 3.0 or Seedance 2.0 Mini, the face gets distorted...
Does this mean I shouldn't touch things that are only halfway decent...?
The creation continues~

• #AIUtilization

• #GenerativeAI

• #ChatGPT

• #AIVideo

• #VideoGenerationAI

• #ForBeginners

• #noteSkiVideoContest

• #AIComparison