[PolloAI PR] Verification! What happens when you change video AI models using the same image and prompt? 🎬

Hello, this is Minato (๑╹ω╹๑ )

Today is a bit of a special episode.
※This article is a PR piece created at the request of Pollo AI. I am writing my honest impressions after actually using it.

I'm always just doing illustrations, but I've actually always wanted to "make my character move" ( ᎔˘꒳˘᎔)…💭

In case you're wondering what Pollo AI is, it's a hub-type platform where you can switch between over 50 different video generation models within a single service.

With just this one app, famous models you've heard of like Veo, Sora, and Kling are all lined up, and its biggest strength is being able to "animate the same image with various models and compare them."

This time, I used the "Image to Video" feature to try out the top two models that were listed as popular!✨️

My thinking is that you can't go wrong if you start with the popular ones😂

I used Seedance and Pollo 3.0.

Almost the same prompt, same single image.

Let's see what's different when you only change the model ( °▽° )

First, let's start with Seedance.

I cut 2 seconds because it was overlapping m(_ _)m
No, but this is just one still image, right? ( '-' )

Even though I only used one still image as the base, the character is moving naturally.

The thing I felt most after using Seedance was its high level of consistency\(｀°∀°)/

Even while moving, the character's face and atmosphere don't easily break down, so there's a sense of relief that "Oh, they're still moving as this character" ( ᎔˘꒳˘᎔)…💭

Seedance seems to be a model that excels at maintaining the consistency of the subject and style within the video, so my experience matched the official description, which was convincing✨️

That said, the credit consumption is higher than Pollo 3.0🥹

This is a trade-off with quality, I suppose🤔

Also, I was quietly surprised that the character limit for prompts is quite generous.

As I will mention later, longer prompts that hit the limit in Pollo 3.0 were fully accepted in Seedance ⸜(*˙꒳˙*)⸝

This might be a welcome point for those who like to specify their direction in detail! ✨️

Next is Pollo 3.0.

The quality here was also unbelievable for coming from a single image (≧∀≦)
The voice is also cute ((o(｡･ω･｡)o))

Compared to Seedance, it consumes fewer credits, so its strength is being able to try it out multiple times without worry.

However, it felt like I needed to provide slightly more detailed instructions for that reason.

Since the interpretation can easily drift with vague prompts, it will respond properly if you carefully write out 'how you want it to move'.

And one more thing, the character limit for prompts is stricter than Seedance.

When I entered the same prompt, it told me 'limit reached' halfway through (ﾉ∀｀lol)

It tests your ability to summarize the main points concisely (ﾉ∀｀lol)

And now, here is the honest corner.

When I entered the exact same prompt as Seedance into Pollo 3.0, this happened.
The result of using the exact same prompt in Pollo 3.0 ⬇️

For some reason, a hand is sticking out lol
Also, the hairstyle is... 🥹

This is completely subjective, but a hand just popped out from a position I didn't intend ( '-' )

But rather than calling this a failure, it's a lesson that 'each model interprets prompts differently' (;´･ω･)

Instead of just reusing a prompt that worked well in Seedance, I need to write it in a way that is tailored for Pollo 3.0. 💡

It was the same as writing different prompts for niji and ChatGPT when generating illustrations ꉂꉂ(ˊᗜˋ*)

In the end, video AI is also a dialogue with the prompt ( ᎔˘꒳˘᎔)

To summarize this verification:

Seedance has higher credit costs, but it has strong consistency and accepts longer prompts. Use it when you want to create that one special piece.

Pollo 3.0 is light on credits and easy to run, but you need to be detailed and careful with your instructions. The character limit is quite strict, so you need to condense your points.

I felt that the right answer isn't about which one is better, but rather using them differently depending on the purpose.

And the thing is, you can switch models with just one tap.

It would be a huge hassle to subscribe to and compare different services, but Pollo AI lets you change just the model on the same screen and regenerate, so verification was incredibly easy.

So this is what a hub-style service means, I see (๑╹ω╹๑ )

To be honest...

At first, I thought, 'I'll connect proper videos to make a work and post it.'

But video generation is harder than I imagined.

Unlike illustrations, there's a flow of time, so there's suddenly so much more to think about.

I'd heard that 'video generation melts time away,' and it was true lol

Before I knew it, hours had passed _( ꜆꒷-ࡇ-꒦)꜆

So this time, instead of overreaching, I presented this as a verification of 'how far a single image can move' and 'how much prompting is needed.'

I've grasped the general usage, so next I want to try something like a battle or isekai theme (ง •̀_•́)ง

I'll show you a peek of what I made 👀

Pollo AI, if I get another project, I'd love to show you a longer video work next time ( ᎔˘꒳˘᎔)

My video challenge, this is how the first attempt went 🍀* ゚

I'd be happy if you could watch over my growth from here (*´艸`)

You can check out the Pollo AI I introduced this time here 👇

Pollo AI

I felt that Pollo AI is a very convenient service for 'people who are unsure which video AI to use' or 'people who want to try out the differences between models using the same material' ⸜(*˙꒳˙*)⸝

That's all for today ✨️

Shio-chan's merchandise is also on sale ╰(´︶`)╯♡
I'm updating it as I go ♪

Also, I'm looking for people for my magazine, so feel free to reach out! (≧∀≦)

👇 My AI illustration management magazine 👇
Minato and Everyone's AI Illustration Gallery 🎨 | Minato Shio

I would be so happy and encouraged if you could like, comment, or follow! ╰(´︶`)╯♡

See you later! (๑╹ω╹๑ )ﾉ

#PR #AIIllustration #PolloAI #VideoGenerationAI #Akarium #ChatGPTCreativeClub