I wanted a real ad for my coloring book, Cozy Gardens, and most of us don't have a studio to shoot one. So I gave one plain product photo and the same prompt to two AI image models on Higgsfield. Round one is Nano Banana Pro, round two is GPT Image 2.6, and round three puts a headline on both that has to be spelled right. Then I turned the winning picture into a 10-second video ad with Seedance 2.5, sound and all.
Every model here is on one Higgsfield plan with one credit balance, and I show what the whole test cost me. Which one would you have picked? Tell me in the comments.
⏱ Chapters 0:00 Which AI makes a product ad I'd actually use? 1:38 Round 1: What can Nano Banana Pro do with one product photo? 3:09 Round 2: Can GPT Image 2.6 beat it? 4:29 Round 3: Which AI can put a headline on an ad and spell it right? 6:56 Can Seedance 2.5 turn the winning picture into a video ad? 9:26 What did this whole test cost on Higgsfield? 10:21 Which Higgsfield plan would I get? 10:38 So which AI won?
📝 My prompts
Ad picture (rounds 1 and 2): Use the attached photo as the exact product: a printed coloring book. Keep the cover art and title exactly as they are, no new words. Make a scroll-stopping Instagram ad: the book lies on a warm wooden table next to a cup of coffee and a few colored pencils, soft morning window light, shallow depth of field, cozy and calm. Leave clean empty space at the top for a headline. Photorealistic, no extra text, no logos.
Headline (round 3): Use the attached photo as the exact product: a printed coloring book. Keep the cover art and title exactly as they are. Make a scroll-stopping Instagram ad: the book lies on a warm wooden table next to a cup of coffee and a few colored pencils, soft morning window light, cozy and calm. Across the empty space at the top, add this headline in clean bold letters, spelled exactly: Put down your phone. Pick up a pencil. No other new text, no logos.
Video (Seedance 2.5, 10s, 720p): Slow camera push-in on the coloring book. Steam rises from the coffee, morning sunlight drifts across the table, and a hand reaches in and picks up a colored pencil. Keep the book cover, the title and the headline exactly as they are. Calm, cozy, quiet morning sounds.
Higgsfield · AI video · AI image generation · AI filmmaking · Free AI video generator · Cheapest AI platform
🙏 Faith + Money: "In all toil there is profit, but mere talk tends only to poverty." (Proverbs 14:23 ESV)
⚠️ Transparency: Some links may be affiliates; I only recommend tools I'd use myself.
Same product, same prompt, two of the biggest AI image models.
One of these ads I'd actually run and one I wouldn't.
Let's find out which is which.
If you sell anything online, a book, a print,
a candle, a digital product, you need ad pictures
and you need a lot of them.
Most of us don't have a video studio
and a photographer and a designer and all that.
So it's really hard for us to put those together.
So I want to know which AI can take one plain photo of my product
and turn it into an ad
that I'd actually want to post.
And the product is my own coloring book.
It's called Cozy Gardens.
Isn't that nice?
Doesn't the name even make you wanna get that coloring book?
But I made this coloring book with AI for an earlier video.
It's live on Amazon now.
I picked it on purpose because it's hard.
The cover has a drawing, a title in fancy letters,
and a subtitle. If the AI changes any of that,
if it's not my book anymore, then this ad is useless.
This video is sponsored by Higgs Field.
I'm doing the whole test on their website
because they have the big image models
and the big video models all in one place.
with one login so it's really easy to try different things
and i don't have to pay for
three separate subscriptions to run one test both models
are going to get the same photo the same
prompt and we're just going to compare them one
quick thing before we start higgs field is a
website you use on your computer there's no mobile app
so use the link in my description and do it
from a computer this is the image page
on higgs field you type in what you want in a prompt box
You pick which AI model makes it.
You can see I've already been doing
a whole bunch of testing on it for different things.
But first off, we are gonna try Nano Banana Pro.
This is Google's flagship image generation model,
so it's the best that Google has to offer.
We're gonna start with that.
It's really known for editing photos that you already have,
which is exactly what I'm doing here, so this'll be great.
I already uploaded a photo of my book,
so I'm just gonna pick it.
That's the plain photo of the cover, nothing fancy.
So here's my prompt.
I'm telling it the photo's the exact product,
keep the cover, the title exactly the same.
And then I describe the ad that I want.
So a wooden table, a cup of coffee,
some colored pencils, morning light,
and an empty space at the top
for a headline that I'll put later.
And the most important line is that first one.
Keep the cover art and title exactly as they are, no new words.
So I want this to look exactly like the product I'm actually selling.
So we're going to make this image 16 by 9.
Instead of 1K, we're going to make it 2K.
So we're going to make the resolution a little bit better.
And before I generate anything,
this tells me how many credits it's going to cost and
image generation is very, very cheap.
You can make a lot of images. And actually right now
with the plan that I have on Higgs field,
I have unlimited nano banana pro right now.
So I'm going to turn that on. So this
actually won't use any credits at all,
but even if I didn't have that, it would just be two credits.
So here's nano banana pros
version of this ad, and this looks really good.
So it's got the light, the coffee, the pencils.
I mean, this makes me want to sit down and start coloring this.
I think it did a really good job, and you can look at the cover.
That is exactly the cover that I uploaded.
So that's exactly what my coloring book looks like.
So for round two, we're going to leave everything the same,
but we're going to use GPT Image 2.6 Sunburst.
So this is OpenAI's version.
We'll try that.
Sunburst is known for handling long prompts really well
and paying close attention to detail,
so I think this will be a good test.
This one would only cost one credit,
but I also have unlimited on Sunburst,
so we'll go ahead and generate this one.
So when I realized that I accidentally left it
on auto instead of doing 16x9 when it started generating,
so we'll switch it to 16x9 and we'll generate another one.
It's free, so we'll generate it.
All right, this came back
and it's a different take than Nano Banana had,
but honestly I think it looks just about as good.
So the cover remained the same, it has the pencils.
The pencils are a little bit crooked.
That looks crooked, right?
Yeah, that's not a straight pencil.
And then it's a little weird.
The book doesn't totally look like it's on top of the table here.
So again, if we go back to Nano Banana,
I do like that GPT image put this little pattern in the cappuccino.
That's nice.
But honestly, I'm going to give Nano Banana the win for this one
because these pencils are straight.
It just looks, it's obviously on the table.
It did better with the shadow, I think.
Let me know in the comments, which do you think did better?
GPT image or nano banana.
Okay, now we're going to work on a headline.
So ad usually needs words on it.
In text is where AI pictures always used to fail.
You'd get spelling errors like a toddler wrote it
or like an alien wrote it or something.
So round three, I'm going to ask both models
to put the same headlines on the ad
and I'll check the spelling, make sure it all looks good.
So we'll start with sunburst since we're already on that.
So all the same settings, but the prompt now
we're going to say put a headline in there.
And the headline that we decided on
was put down your phone, pick up a pencil.
That's pretty good, isn't it?
I think that's inspiring.
All right, and again, still on unlimited,
so let's go sunburst.
And you know what, we don't need to wait.
Again, the beauty of AI is you can start something
and then work on something else.
So let's switch this one to Nano Banana Pro,
and we'll do the exact same prompt,
and we'll get going on both of them at the same time.
Multitasking.
All right, first the Nano Banana one.
Put down your phone, pick up a pencil.
It changed the ad just a little bit from what it did
when I didn't say the headline.
But honestly, I think this looks pretty good.
I like how it's on there.
They picked a good font, picked a good color that you can read.
And did it spell it correctly?
Put down your phone, pick up a pencil.
Yes, it did spell it correctly.
So AI has come a long ways.
Congratulations, AI.
All right. Then here's Sunburst's version.
This, I actually think, looks maybe a little
bit better than the ad. The book is clearly on the table now.
One thing is this is a weird table
that goes up against the window and the wall in the back.
I think that looks a little strange.
They did the pattern again in the drink, though, which I like.
Um, the words too, I don't, they're fine.
And then put down your phone, pick up a pencil. So
it did spell it all correctly,
but I don't think it stands out quite as much as a Gemini did as
nano banana did. So actually I'm going to probably say,
let me know in the comments if you think I'm
wrong, but I'm going to put nano banana
as the winner for both the ad itself and the ad with
a headline. Um, but let me know in the comments
if you think differently, but that's what I see.
So consensus on the image generation, and stay tuned,
we're going to do video as well, a video
ad. But what I was mainly wanting to look at,
all the versions, Sunburst and Nano Banana, they all
kept the cover the way that it should,
like that looks the same as my book. Honestly, they all did
a pretty good job. The headline was great on both,
but I do pick Nano Banana as the winner for just
looking a little bit better, a little bit more realistic.
All right, now let's test video.
Because a picture is good, but a video ad usually gets more attention.
So let's take the winner here and let's do more with it.
And I actually don't even need to reprompt.
Higgs field has a turn to video right here.
So I'm just going to click that.
Now it takes me over to the video generation.
You can see I've been messing around with that a little bit.
And I do really like in Higgs field that you have images
and videos just all together,
along with a whole bunch of other stuff.
I could make probably 50 videos about Higgs field.
So we're starting with that nano banana image that was a winner
and then the prompt that we're giving it
is just basically what I'm envisioning in this video.
So slow camera push in on the coloring book,
steam rising from the coffee, morning sunlight drifts across the table,
a hand reaches in, picks up a colored pencil, keep the book cover,
the title and the headline exactly as they are, calm, cozy,
quiet morning sounds. For the model we're going to use Seed Dance 2.5.
In my opinion it's the best video generating model right now.
I like Seed Dance too.
It does the video and the sound together.
Some video models will just do video.
You have to do the sound separately
and try to match them up.
Way easier if I can just put the sound
as part of the video.
Quality, just to save money,
we don't need 1080p and you can see it saves quite a few credits going
down to 720p. So tip for you guys, you probably don't need 4K or 1080.
720p is probably fine and
it's going to be way fewer credits.
For duration, Seed Dance can go all the way up to 30, but we
don't need that. I mean, if you look at what I asked the video to do,
that 10 seconds is plenty.
And actually on a paid plan,
I get unlimited seed dance until November 25th.
So Higgs Field does a lot of just unlimited
for you to try out the tools and mess around
with them.
So I'll turn on unlimited mode.
So it'd be 70 credits without that.
But actually, I get this for free.
We'll go ahead and generate.
Video takes a little bit longer than image does,
but we'll check back here in a second.
All right.
So this ad is done.
Let's check it out.
Looks pretty good. The hand was pretty good.
Remember, AI used to always mess up hands.
And I was actually nervous about them picking up a pencil,
but I think that looks pretty good.
Although they're drawing on the cover,
but I think it's probably all right.
That's 10 seconds with sound from one plain photo.
And I think I could put that up today, potentially as an ad.
So what did all this cost me?
That's the big question with AI generation is how much is this
going to cost? Yeah. AI can do cool things.
Is it going to cost, you know, tons of money on my
paid account with Higgs field, nano banana pro and C dance.
We're all on unlimited. So those were
free. The GBT image pictures were also free.
So really I spent actually nothing.
To make what you saw here today.
If you don't have unlimited,
the generate button shows you how much it'll cost
before you click, just like we saw.
But I think everything we did today, not very many credits.
It's getting a lot cheaper to do something like this.
And of course compare that to hiring a product photographer.
You'd pay for the shoot,
then you'd wait a few days for the pictures.
Then you'd still need to pay someone to make the video.
Here I got four ad pictures and a video clip
in a few minutes.
And I can make 10 more versions tonight.
Especially while we're on Unlimited,
just go crazy with it and I'll make a whole bunch of versions.
Higgsfield is doing a Black Friday sale.
If you go through my link in the description,
you can get up to 60% off all Higgsfield plans.
And the plan that I would get is the Ultra plan.
You get 6,000 credits per month.
You get all that unlimited stuff that you saw me do today.
Link is in the description and the pinned comment.
So which AI makes a product ad that
I would actually use? I think I'm giving
the winner for the image to Nano Banana.
Seed Dance 2.5 is amazing.
I use it every day for tons of stuff.
And you don't have to pick just one.
You can hop on here, follow exactly what I did today.
Do your own comparison, do your own tests.
Maybe you make a YouTube video
where you compare different AI image generation models.
So appreciate you guys so much for watching.
Please do like and subscribe.
The Currents of YouTubers, we appreciate it.
Thank you very much.
God bless you all and have a great rest of your day.