In this Dreamina Canvas AI Design tutorial, I upload a 15-second coffee commercial, use "Analyze this video" to break it down shot by shot, and build a new ad for a different café with Seedance 2.0. You'll see how the Canvas agent turns the analysis into prompts, how reference images keep the location and the character consistent, how to redo a single shot, and how I finish the edit in CapCut.
Chapters: 0:00 CapCut Pro giveaway: up to 10 free Seedance 2.0 generations 0:38 Why Seedance 2.0 on Dreamina (ByteDance, CapCut, Seedance) 0:55 The plan: rebuild a coffee ad for a different café 1:35 Dreamina Canvas: "Analyze this video" breaks the ad down 2:15 What the analysis gives you: four beats and a camera style 2:36 The Canvas agent writes three Seedance 2.0 prompts 3:04 Generating the shots: video nodes and Omni reference 3:40 Redoing one shot with Frame capture to keep the same character 4:33 Finishing in CapCut and the final ad 5:12 Using this workflow for your own ads, and what else Canvas can do 5:45 How to enter the giveaway
Dreamina, CapCut and Seedance are all part of ByteDance.
This video explains how to use Dreamina, ByteDance's AI creation platform, to reverse-engineer and recreate professional commercials. The creator demonstrates a workflow that involves analyzing an existing ad's structure and camera movements to build a new, consistent commercial for a different brand, ultimately finalizing the edit in CapCut.
Hi everyone, if you have CapCut Pro or a higher plan,
Dreamina has a giveaway running right now,
and everyone who enters wins.
To enter, open the link in the description. The draw pops up on its
own. You click sign in to reveal,
login in Dreamina web with the same account you used for CapCut,
and you see your result. You get 3 things,
You can get up to 98% off a month of Dreamina Basic,
and the amount depends on your draw.
and if you're new to Dreamina, that
month can cost as little as 29 cents.
You get up to 10 free Seedance 2.0 video generations
at 720p with no limit on video length, and
you get extra bonus credits.
The free generations are what got my attention.
Seedance 2.0 gets expensive quickly when you're still testing ideas.
Dreamina is ByteDance AI creation platform.
ByteDance also make CapCut and Seedance itself.
So you are using the model where it was built, at full strength.
So here's what I spent those generations on.
I take a coffee commercial I like,
I lay Dreamina canvas, take it apart and use what it
finds to build a new ad for a different cafe.
You start with an ad you like, mine it's a 15 second coffee commercial.
One note, use a video you made yourself or one that's free to use.
The ad I need is for another cafe,
Olivo Coffee, so I can copy the look.
What I want
is the way it's
put together.
In Dreamina web, the create page has two modes,
generate is the quick one, a prompt box for
a single image or video, canva is for a project like this,
where every step sits on a board
as a node I can go back and change.
So I open a new canvas, upload the clip and choose analyze this video.
I don't write a prompt for this,
the agents split the ad into time segments on its own
And for each one it describes what's in the frame
and how the camera moves.
Seconds 2-7, for example, are a fast run of close-ups,
steam, milk, latte art, the lid
going on.
Above that it sums up the visual style and even the sound.
Those are details that are easy to miss when you just watch an ad.
Here's why that's useful to me.
The breakdown shows the ad is built in 4 bits,
the room, the coffee being made, the customer
enjoying it and a wide shot to close.
It also named the camera style,
which is steady framing and tight close ups on the craft.
I can reuse all of that.
I keep the beats and the camera and I change everything you see.
Olive already has its look.
I upload one image of the cafe to the canvas
and it becomes the location for every shot.
Then I ask the agent for three prompts,
one per shot, right under the structure and camera
style of the commercial it just analyzed.
The prompts arrive as three text notes
and you can see the analysis inside them.
Each one opens with the same pacing and camera notes,
then lays out the shot second by second.
And I can edit any of it right there on the canvas,
before I generate anything.
Now the video. Each text node connects to its own video node,
and the cafe image connects to all three.
I pick Gen 2.0, 10 seconds, landscape.
One setting matters here, and that's Omni Reference.
With it, the cafe image works as a location, not as the first frame,
so a close-up can start right on the cup.
The node can also run several takes at once.
For this ad, one per shot is enough.
The first and the last shots come out right on the first try.
It's the same cafe in both
and the ad even closes on a wide shot of the room the way
the original does.
The middle shot is the one I redid.
The hand picking up the cup came out in slow motion
and I didn't like it.
Since I was redoing it anyway, I changed two things.
I rewrote the shot to note
so she gets her coffee and takes a first sip.
And I used frame capture on the last clip to save a frame of the woman.
Then connect it to the video node as a second reference.
That way is the same person in both shots.
Then I regenerate only this one.
The other
two stayed
as they were.
Now it's what I wanted.
It's clearly her, down to the jacket.
And she's holding the cup she carries out the door.
That sip with her eyes closed is the face of someone
who hasn't opened her email yet.
It's also the customer beat from the original ad.
So mine now has all four.
That was the hard part.
From here I download the three clips and finish in CapCut.
I trim the dead moments,
add the track from CapCut's music library and close on the
Olivo logo.
Let's watch the entire ad.
So if you usually
open CapCut
without knowing
what to make,
start one
step earlier.
Break down an ad you like in Dreamina,
build your own shots on its structure, then finish
in CapCut.
For this ad that meant 3 shots and 1 redo.
And the reference doesn't have to be a coffee ad.
It could be a TV style commercial, for a product video,
for a client or for your own products.
This only scratches the surface too.
Canvas also has a timeline for rough cuts,
motion capture and a 3D stage for planning your shots
I didn't want to scare you in the first video,
that's what the next one is for.
If you are on CapCut Pro or higher,
the giveaway link is in the description.
Sign in with your CapCut account
and you could get up to 98% off Dreamina Basic, up to 10
free Seedens 2.0 generations and extra bonus credits.
Tell me in the comments what your draw gave you.
I'm Dani, take care and I'll see you in the next one.