13:44Pat Simmons · Tutorial
Fable 5.1 + Seedance 2.5 Just Unlocked Vox-Style Motion Graphics
One prompt file sent an autonomous coding agent off to research, script, art-direct, and render a full Vox-style explainer video by itself.
September 3rdThe argument in one line.
Seedance 2.5's new text-rendering and 30-second generations make AI motion graphics production-ready, but a Claude Code skill that writes the prompts and five CapCut effects layered on the output are what make the result pass as professionally animated.
Read if. Skip if.
READ IF YOU ARE…
- A motion designer or video editor curious whether AI video models can now handle on-screen text well enough to use in real client work.
- Someone running a Claude Code or Claude Chat subscription who wants a repeatable prompt-writing workflow instead of guessing at AI video prompts.
- A solo creator or small agency building ads, explainers, or app promos who wants a fast route to Vox-style or Apple-style motion graphics.
SKIP IF…
- You need a completely free workflow — this depends on a Higgsfield.ai subscription for Seedance 2.5 access and a Claude subscription for prompt generation.
- You're looking for hand-keyframed, fully custom animation — every example here is generated from a single text prompt, not built shot by shot.
The full version, fast.
Seedance 2.5 is the first AI video model that renders clean, intact on-screen text and holds a single generation for up to 30 seconds, which finally makes it usable for real motion graphics. The workflow: pull a style reference from Pinterest, hand it to a custom Claude skill that writes a detailed Seedance prompt, paste that into Higgsfield.ai, and generate. The video demonstrates it across simple text animations, a full brand explainer from one prompt, a real product ad recreated for a different product, and a Vox-style collage explainer, then finishes each render in CapCut with five layered effects, shutter stutter, dual tone, chromatic blur with a sharp-focus mask, radial edge blur, and grain, to strip the AI polish and make it read as hand-animated.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 00:30
01 · Cold open
A rapid-fire showcase of finished Seedance 2.5 motion graphics (a collage book, neon text, a mountain map reveal) sets up the claim that new text rendering and longer generations changed what's possible.
00:30 – 01:09
02 · What this video covers
Matt outlines the three styles he'll demonstrate: simple text-based motion graphics, complex multi-scene animations from one prompt, and Vox-inspired collage explainers.
01:09 – 02:00
03 · Step 1: install the Claude skill
He walks through downloading a custom-built Claude skill (zip file linked in the description) and installing it via Claude's Settings > Skills > Add so Claude writes richer Seedance 2.5 prompts.
02:00 – 06:18
04 · Simple text-based motion graphics: prompt bar example
A Pinterest 'aesthetic UI' reference image is dragged into Claude along with a written brief; Claude asks clarifying questions, generates a detailed multi-shot Seedance prompt, and the resulting prompt-bar animation is generated in Higgsfield.ai at 10 seconds, 1080p.
06:18 – 07:05
05 · Example 2: search bar animation
A second Pinterest reference and near-identical Claude prompt produce a Google-style search bar expanding into on-screen text, following the same drag-image-then-prompt loop as the first example.
07:05 – 09:14
06 · Example 3: typography-led product ad, and fixing broken text
A Pinterest video (not just an image) is downloaded and analyzed frame by frame by Claude; the first Seedance generation looks strong but the text jumbles halfway through, so re-uploading the failed clip to the same Claude chat produces a corrected prompt.
09:14 – 11:26
07 · Complex multi-scene animations: the Jasper brand ad
Instead of copying a reference video's look, Claude is told to borrow only its camera motion and pacing, then apply that motion to an invented brand ('Jasper') and its own visual identity, producing a 28-second single-prompt brand explainer.
11:26 – 12:12
08 · Real product test: iPhone 17 Pro remake
A low-quality real iPhone ad is uploaded to Claude with instructions to analyze it and write a matching Seedance prompt that keeps text fully intact, producing a sharper remake that integrates a physical product into legible on-screen text.
12:12 – 13:13
09 · Swapping the product: Bell Pepper Pro
The same visual system is redirected onto a different product (three bell peppers) by asking Claude to keep the styling and motion but rewrite the story and remove all Apple branding, showing the technique works on any product, not just the one in the reference.
13:13 – 14:43
10 · Vox-inspired collage explainer
Claude first writes a 20-second script about the Voynich manuscript, then a Pinterest collage-style reference image is used to turn that script into a full collage-animation Seedance prompt, producing a Vox-style explainer in about five minutes.
14:43 – 20:34
11 · CapCut finishing pass
The finished Seedance render is layered with five CapCut effects, FPS Lag (stop-motion stutter), Dual Tone (texture), Chrome Blur with a reversed circle mask (fake depth of field), Radial Blur (warped edges), and a masked adjustment layer plus grain, to strip the AI look; he notes which effects suit which style (Dual Tone for collage/stop-motion, not clean product ads).
Lines worth screenshotting.
- Seedance 2.5 is the first AI video model that can render new on-screen text within a scene rather than just preserving text already on a product.
- A single Seedance 2.5 generation can run up to 30 seconds and hold multiple distinct scenes, where earlier models needed clips stitched together.
- Feeding Claude a screenshot of a UI or design style you like, plus a written brief, produces a far more detailed Seedance prompt than writing the prompt from scratch.
- Uploading a reference video to Claude and asking it to analyze it frame by frame produces a usably accurate motion breakdown; without saying 'frame by frame,' it only samples about 10% of the footage.
- You can reuse an AI ad's exact camera motion and pacing for an unrelated brand by telling Claude the reference video is for motion principles only, not for color or styling.
- Feeding a real, blurry product ad into Claude and asking it to write a matching Seedance prompt can output a sharper, cleaner remake of the same ad.
- A generation that looks great for the first half can degrade into jumbled text later on; re-uploading the failed clip to the same Claude chat and flagging the issue produces a corrected prompt without starting over.
- Layering CapCut's FPS Lag effect at around 40% turns smooth AI motion into a stop-motion stutter that visually signals hand animation rather than AI generation.
- A reversed circle mask on the Chrome Blur effect keeps the subject sharp while blurring everything outside the mask, creating an artificial depth-of-field point of focus.
- Dual Tone works for stop-motion and collage-style animations but visibly degrades clean product renders like an Apple-style ad.
- Five CapCut effects, FPS Lag, Dual Tone, Chrome Blur, Radial Blur, and a masked adjustment layer with grain, take about five minutes to apply and are what separate an obviously-AI clip from one that reads as professionally animated.
What separates AI video from usable video.
PROMPT + FINISH
A good AI video model only gets you halfway: the reader gets there by pairing a structured, reference-driven prompt process with a deliberate finishing pass that removes the telltale AI polish.
03Step 1: install the Claude skill
- A packaged Claude skill built from testing what phrasing a specific AI video model responds to writes far richer prompts than typing requests from scratch.
- Installing a custom skill is a one-time step: download the zip, then add it under Claude's Settings, Skills, Add, so every later prompt in that chat inherits it.
04Simple text-based motion graphics: prompt bar example
- Pulling a single reference image from Pinterest and describing the desired motion in plain language is enough for Claude to draft a full multi-shot prompt, including camera moves and typed-out on-screen copy.
05Example 2: search bar animation
- The same reference-plus-brief pattern reproduces reliably on a second, unrelated animation, which is the real test of whether a workflow is repeatable rather than a one-off lucky prompt.
06Example 3: typography-led product ad, and fixing broken text
- Telling Claude to analyze a reference video frame by frame is necessary; without that instruction it samples roughly a tenth of the footage and misses the actual motion and pacing.
- When a generation degrades partway through, re-uploading the failed output to the same Claude conversation and describing the defect produces a corrected prompt without redoing the analysis from scratch.
07Complex multi-scene animations: the Jasper brand ad
- A reference video can be used for motion and pacing alone, decoupled from its color and branding, by explicitly telling Claude which parts of the reference to ignore.
- A single prompt can now carry a full brand identity, logo, product description, visual system, into a 28-second multi-scene animation, a task that used to require manual keyframing.
08Real product test: iPhone 17 Pro remake
- Uploading a real, low-quality product ad and explicitly asking for text to stay completely intact in the prompt produces a sharper remake with legible on-screen copy, fixing what older AI video models couldn't do.
09Swapping the product: Bell Pepper Pro
- The visual system and motion of a generated ad can be transplanted onto a completely different product by keeping the prompt's styling instructions and swapping only the subject and branding.
10Vox-inspired collage explainer
- Writing the script first as its own request, then converting it into a visual prompt, keeps the narration and the animation from fighting each other.
- A collage-style explainer that would take hours to hand-animate can be generated in about five minutes once the script and a reference image are ready.
11CapCut finishing pass
- A stack of five CapCut effects, frame-rate stutter, a texture layer, a masked chromatic blur, edge warp, and grain, takes about five minutes and is what makes an AI clip read as professionally animated.
- Not every finishing effect fits every style: a texture built for stop-motion and collage actively hurts a clean, premium product ad, so the pass has to match the source material's aesthetic.
Terms worth knowing.
- Seedance 2.5
- A text-to-video AI model built for motion graphics generation, notable for rendering legible new text mid-scene and generating up to 30-second clips.
- Higgsfield.ai
- A web platform used to run Seedance 2.5 generations at 1080p, with reference-image upload and duration controls.
- Claude skill
- A packaged set of custom instructions installed into Claude that shapes how it writes prompts for a specific tool, here tuned to Seedance 2.5's preferred prompt style.
- FPS Lag
- A CapCut effect that lowers the apparent frame rate of a clip to create a choppy, stop-motion-style stutter.
- Dual Tone
- A CapCut effect that overlays a stylized two-color texture onto footage, commonly used on stop-motion or collage animation.
- Chrome Blur
- A CapCut effect combining blur and chromatic aberration, used here with a reversed circle mask to fake a sharp focal point.
- Radial Blur
- A CapCut effect that warps and blurs the outer edges of the frame while leaving the center sharp.
- Feathering
- The CapCut mask control that softens the hard edge of a mask so an effect fades in gradually instead of cutting off sharply.
Things they pointed at.
01:09toolClaude
02:27toolPinterest
15:03toolCapCut
Lines you could clip.
00:00
“Seed Dance 2.5 is able to create consistent professional motion graphics and this is for a number of different reasons.”
08:00
“If you don't mention frame by frame, Claude tends to analyze around 10% of it, which is not what we want.”
13:00
“This shows you that you can use your product. You can use a client's product.”
15:03
“All of these effects put together in CapCut take five minutes to place on top of your video, but add extra depth and that professional feel to your video.”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
metaphor
Seed Dance 2 .5 is able to create consistent professional motion graphics and this is for a number of different reasons. The first is we finally have intact text. We get 30 second long generations and we have a system that understands our prompts a whole lot better.
Despite this, I'm going to be showing you a little bit later how to take these motion graphics and really set them apart by doing some tweaks in our editing software that help them to feel professionally animated. In this video, I'm covering a few different styles of motion graphics. The first is simple text -based motion graphics like you're seeing on screen right now.
The next is complex multi -scene animations from one single prompt. And finally, Vox -inspired explainers, all done in CDONS 2 .5. now there are a couple of crucial steps that i recommend that you follow the reason this is is because the worst thing is seeing a video like you're watching right now seeing the potential of what cdons 2 .5 can make you go and test it yourself you waste a bunch of credits and you're not able to get a professional motion graphic so follow the steps and i'll walk you through exactly how i wished someone walked me through it when i was starting out so the first practical thing that you're going to do is actually to install a skill into claude This is going to help create rich, in -depth prompts that we can use to create better generations.
Now, I've custom made a skill just for you based off my custom testing and the style of prompt that SeedDance 2 .5 likes. Now, installing this skill is super easy. There is a download link in the description.
All you need to do is click on that download link and you'll have access to the SeedDance 2 .5 zip file. Once you've downloaded that zip file, you're going to go into your cloud, click in the bottom left, go to settings. go down to skills, then click add, upload skill, click on this folder icon, and you'll see that that zip file is in your downloads.
Go ahead and click on that and click open. once again this gives claude all the information it needs to create really rich deep prompts for sedans 2 .5 our first style of motion graphics is those simple and text -based motion graphics as i mentioned prior all video generation models really struggled with text some of them could take text off of a product but they struggled to generate new text within the scene this changes in cdance 2 .5 enabling a whole new style of generation motion graphics that used to take days to edit now can be done in minutes for this first example i'm going to be generating this light style prompt bar step one is to go to a site like pinterest and in the search bar we can search for aesthetic ui this is going to give us a whole bunch of beautiful user interface icons and graphics that we can use to inform the type of styling that we want in our prompt bar.
When you've found an image that you like, simply click on that, go to these three lines, and download this image to your computer. On screen right now is the image that I downloaded from Pinterest, and you can see that there's a correlation between the styling and the styling of our final motion graphic. What we're then going to do is go to Claude, and you can either use the chat section or the code section.
Personally, I like using the code section. It just feels a little bit more robust. But if you like using chat, go ahead and use that as well.
It's going to create no difference in the quality of our prompt. What we're going to do is directly drag our reference image of the styling into Claude. Then type your prompt.
To start our prompt. I wrote use forward slash cdance 2 .5. This is going to make sure that we use that skill that you just installed.
I've also said I've attached an image with the color and visual design style I'd like for an animation or motion graphic using cdance 2 .5. Think light aesthetic, clean background and pastel type colors like the reference. Now this is where I actually informed Claude exactly what I wanted in the motion graphic and you don't necessarily have to do this.
This just helped us get to that final result a little bit quicker. The animation is 10 seconds long, multi -shot. We start on the prompt bar close -up, it expands.
We realistically type in the prompt bar, motion graphics made easy, how? Camera then cuts to a close -up on the send button. A realistic mouse click send, and cool motion graphics pop up all around.
All clean, smooth motion, like a professional motion designer. and finally i put ask me questions to lock this in on the first attempt let's go ahead and click send if you'd like to follow along and have copyable versions of these prompts i've created a document where all of the prompts that i used in this video will be placed in you can then just copy them from that document and place them into your claude and have the exact same generations that i've made after waiting around 30 seconds claude is going to ask you a couple of questions to clarify the exact direction that you want to follow This is saying I've already created this motion graphic because I have but all you then need to do is click the answers that you like just by simply reading them and once that's done Claude is going to actually start to generate the prompt for your motion graphic.
Once the prompt is done generating what it's going to look like is either an MD file which you can click and have access to the entire prompt or it's simply going to paste that prompt in the direct chat. All we need to do is go ahead and navigate to this copy icon and click copy code and then for the actual video generation click the link in this description which is going to take you to higgsfield .ai we're using higgsfield .ai because it has access to seedons 2 .5 and 1080p generations which is obviously going to create really sharp beautiful looking motion graphics once i'm on the home page i can either click seedons 2 .5 here or navigate to video swipe down and go to seedons 2 .5 what i then did is click add reference and i uploaded that image that we saved from pinterest after a few seconds it was done processing and all i had to do was click that image and then go to our prompt bar and click paste now something to note make sure that your model is set on cdons 2 .5 also go to your duration we obviously explicitly said that we wanted a 10 second generation so make sure this is set to 10 seconds you can change your aspect ratio i'm going to leave this on 16 by 9 and then let's leave this on 1080p all we then need to do is hit generate
After a few minutes of waiting we have our professional motion graphic.
For this next style of simple text -based motion graphic, which looks incredible, I followed a very similar style. I found a reference image on Pinterest, which I really liked. I saved that to my computer and uploaded it into Claude.
I then typed a prompt. Using the image attached as a guide to color and visual style, write a prompt for CDONS 2 .5. My prompt then generated in the actual chat and didn't create that separate file.
All I had to do was click copy. Go back to Higgs field, delete this prompt and delete our reference. Let's add a reference and paste that image that we used to inform Claude.
I then simply pasted the prompt right there, made sure my settings were all correct and clicked generate. We then got this really stunning motion graphic.
For this third example, I wanted to create a typography -led advert for a product of mine. So I went to Pinterest and searched motion graphic with text. You'll see that a bunch of them have time codes in the top left.
That means that they're an actual video. I was inspired and I simply hovered over a couple of them until I found one that I really liked. This one here was the exact one that I liked, so I went ahead and clicked that.
Using this link in the top, we can go ahead and copy that and search for a Pinterest video downloader. Simply by using one of these results, paste your link and download that Pinterest video. Now the great thing about Claude is that when we use Claude Code or Claude Chat, we can drag and drop our video into the chat bar.
Once your video is attached, make sure that you mention to Claude to analyze the video frame by frame. If you don't mention frame by frame, Claude tends to analyze around 10 % of it, which is not what we want because it then doesn't understand the motion and the style of motion graphic that we like. Just like the other ones, what we then need to do is copy that prompt and paste it directly into Higgs field.
Now, I didn't add the video reference for this example, just the prompt, and it understood it perfectly. Now, I must be honest, the first generation that it did looks stunning in the beginning. But as we swipe ahead here, you'll see that around halfway, we start to have jumbled text.
Now I didn't like this, so what I did was I downloaded this video, I re -uploaded it into that Claude chat and made sure that it understood that there was an issue with the text. Claude then remedied the problem, created a new prompt which fixed the text and created this video.
And that's one of the phenomenal things, in my opinion, about working with an AI like Claude. We don't have to necessarily self -remedy. We can just upload the video that failed to Claude and it'll fix it itself.
This next section is all about creating those complex multi -scene motion graphics like you're seeing on screen right now. Go to Pinterest and find a motion graphic style that you really like. Just like we downloaded our other example, we're going to download this video directly to a computer.
But there's a couple of different steps. Inside of Claude, create a new chat. And let's go ahead and drag and drop that reference video that we like.
Now, what I did differently in this prompt was I uploaded that MP4 so Claude could analyze it. But instead of saying that I want to emulate or copy this video, I actually gave Claude a bit of a different brief. I gave Claude a different aesthetic.
I told it that I want smooth motion with dynamic camera moves. And I said I have attached a video for camera and motion principles, not necessarily for color or styling. Now, this is when you get to.
upload your client's logo, upload your own logo, upload a brief of exactly what your product does. You can change anything about this. All we're using the video for, this video reference, is as an example to the camera motion that we like.
I gave Claude a tiny bit of context as to what this product that I made up called Jasper does, and I click send. With the prompt generated, once again, I click copy. I went back to Higgs field, pasted it directly into our prompt bar.
I can see that my animation was 28 seconds long, so I needed to change my duration to 28 seconds, leave all the settings the same, and in one click, it created this.
That blows my mind. Perfect text, it created those dynamic camera moves, it was kind of a singular scene type style, and it understood the motion principles from our example, but we used a complete own aesthetic, our complete own brand guide, and it created that. Now I realize that CDON's 2 .5 was phenomenal for motion graphics and animation, but what about using a physical real world product?
I found this really low quality example of an iPhone advert. you can see here on the screen right now this scene was extremely blurred but what i did was i uploaded this video into claude i said can you analyze this video and write a seedlons 2 .5 prompt for a similar generation make sure that the text stays completely intact with a specific line in the code asking for this claude then ran created this iphone md document with the code i copied the prompt uploaded it into higgs field and it created this What's incredible to me about this is we don't only get that perfect text, but we actually integrated a physical product into that text.
But I took this one step further. In the same Claude chat, I uploaded an image of three bell peppers. I said, I really like the style of the video we generated, but I want to make it about three bell peppers instead.
Make sure all Apple logo and branding is gone and rewrite the story to be about bell peppers. Maintain the visual styling, color changes of the background, etc. Once the prompt was done, I simply copied it like we've been doing.
I added a reference of the ballpapers. This is going to make sure that those colors of the ballpapers stay the same. I pasted in our prompt and I clicked generate.
What this revealed to me was I could generate something that I really liked off an existing product that wasn't necessarily mine. Then I could use that visual styling, the feel, the Apple aesthetic, the motion. And I could change that iPhone or whatever product into my own product, or in this case, ballpapers.
So this shows you that you can use your product. You can use a client's product. And for our third example of the style of motion graphics you can create, we're generating full -scene Vox -inspired explainers.
And this was more than just creating the visuals, and I'll show you why. I went into Claude and I created a new chat. I then asked it to write a 20 second script that'll later be animated all about the Voynich manuscript.
And I click send. While Claude was generating our script, I went back to Pinterest and I searched for collage design. Immediately, I found this collage style animation that I really liked.
I hovered over those buttons and clicked download image. With our script generated, I drag and dropped that image directly into Claude. With my image dragged in, I simply wrote, using the story you just wrote, write a 20 -second collage animation prompt for CDONs 2 .5, maintaining the style of collage and text integrity.
Once my prompt was done generating, I copied that, I uploaded our reference image, pasted that into the prompt bar, and I knew my generation was 20 seconds, so I made sure this was 20 seconds long. I hit generate, and it created this.
To edit this myself, honestly, it probably would have taken me a number of hours. And within five minutes, I had a Vox -style generation that I was actually proud of. But this leads us on to our next step.
Overall, I felt these motion graphics were insane, like I said, but there was a couple of things that I knew we could do to make them feel more customized and professional. That's when moving to CapCut steps in. Now, you can do these steps in your other editing softwares.
I just use CapCut because I know the effects that it has. and they honestly look incredible. So the first thing that I did was I downloaded that Vox style generation directly onto my computer and imported it into CapCut.
Let's go to effects and search for something called FPS. The first result is FPS lag and this is a slow shutter effect which Vox does all the time. It almost makes it seem like we're watching a bit of stop motion.
I'm gonna drag that over my whole video and click play. You can see all of a sudden, if I turn this off, we have smooth motion. And if I turn it on, that motion gets a little choppy.
And I'm going to turn it down to around 40. Now, all of a sudden, we have that stop motion style, which immediately sets this apart from a AI generation. The next thing that we're going to do is search for dual tone.
I'm going to go ahead and drag that first result onto my timeline and extend it over my whole video. I'm also going to turn down speed to around two. And you'll see that if I zoom in, By turning dual tone off, we remove that texture and by turning it on, it now looks like we have a bit of a tone or texture on top of our video.
The next thing I'm gonna search for in my effects is chrome blur. Let's go ahead and drag that first result onto our timeline and extend it once again over our video. Now this creates a really cool chrome effect, but it honestly makes our video look unprofessional.
So what we're gonna do is turn down blur and turn our lateral chromatic aberration to around 52. We're then gonna go to mask. click add mask and create a circle mask let's go ahead and make our circle mask quite a bit bigger by dragging it top and bottom and dragging it left and right what we're then going to do is go to this option right here and click reverse this is going to make sure that everything within the mask is perfectly sharp and everything outside of the mask as you can see here has that chromatic blur we're also going to go to feather and start turning up feather until we get a really smooth effect i think 24 looks perfect let's take a look at what this effect is doing You can see if I drag to the corner here when our effect is on we get that chromatic blur Where this man is slightly blurred and we get that cool aberration if I turn it off You'll see that he's perfectly sharp just by toggling it on and off You can see that we've created a point of focus in our video Where everything on those outer edges is kind of blurred the next effect we're going to apply is something called radial blur Let's go ahead and drag radial blur 2 onto our timeline
and let's quickly take a look at what it's doing you can see that on the outer edges of our video radial blur is creating this cool blur where it almost warps that side of the image now personally i think this is slightly too much so i'm going to go ahead and start turning down our size to around 29 let's go ahead and turn filters off and glow off and we can also decrease the strength slightly if i move to a scene which occupies more of the screen and turn it off look at what it's doing on those corners It's creating a warped effect.
We can go ahead and turn this down slightly less to 27. And it's just subtly, if I zoom in here on my screen, it's just subtly warping that screen, which once again, adds a cool extra effect to your video, which AI most of the times wouldn't be able to do. One of the last things we're gonna do is add a custom adjustment layer in the adjustment tab.
And again, you guessed it, we're gonna fill up our entire timeline. Now go to color wheel and turn down offset. to around this point.
This is totally customizable depending on your video. We're then gonna go to mask, add mask and create a split mask. This creates a straight line which we can drag diagonally and increase the feathering just like that.
I'm also gonna drag this mask down into the corner just so we have a darker area there. I'm also gonna create a second split mask, drag it to the side. and do the same thing let's create some feathering and drag it to the top right by turning this on and off you'll see that we now have areas of dark in our corners once again this creates a stronger central focus in our video and adds a little bit more of contrast and depth to our image the last thing you can do is select on your video go to adjust swipe down and add a couple of points of particles i'm going to go ahead and turn off all of my effects so you can see the particles It's really subtle at this point, but if I go ahead and turn them up, you'll see the effect that it's having.
It's creating some moving grain in our video just like that. Again, I don't really like the effect to be that strong, so I'm gonna make it 10. And again, if I zoom in even more, you'll see that it's just some tiny particles in our video.
All of these effects put together in CapCut take five minutes to place on top of your video, but add extra depth and that professional feel to your video.
Now I should also preface that I wouldn't necessarily put all of these effects on every style of video for this Apple style video I wanted it to look clean and premium So placing dual tone on top of the video honestly just makes it look really bad Dual tone mainly works for stop -motion or collage style animations But if we go to adjust radial blur and chrome blur we can go ahead and copy those effects over And if I turn this on and off you'll see that we're definitely getting that adjust on the top and the bottom just there.
And then if I go to a scene where our phones are coming in from the side, you'll see that we get that cool chromatic aberration and lack of focus on those corners from our radial blur and our chrome blur. And that's how you make those professional looking motion graphics in C -Dance 2 .5 and CapCut. If you want access to that Claude skill, those prompts, or if you want to access higgsfield .ai, go to those links in the description and you'll have everything that you need.
The Hook
The bait, then the rug-pull.
Seedance 2.5 finally renders text inside a generated scene instead of garbling it, and this creator built a Claude skill around that one capability, then ran it through three motion graphics styles and a five-effect CapCut finishing pass to see how far the AI-to-professional gap could close.
Named ideas worth stealing.
01:09list
Three Seedance 2.5 motion graphics styles
- Simple text-based motion graphics
- Complex multi-scene animations from one prompt
- Vox-inspired collage explainers
The video's own structure: three escalating styles of motion graphic, each built with the same Claude-to-Higgsfield prompt loop.
Steal forany AI-video content plan that needs to demonstrate range in one video
15:03list
Five-effect CapCut finishing stack
- FPS Lag (~40)
- Dual Tone (speed ~2)
- Chrome Blur + reversed circle mask (chromatic aberration ~52, feather ~24)
- Radial Blur 2 (size ~27-29)
- Adjustment layer: color wheel offset + two split masks + grain (~10)
The exact stack and rough settings used to take a Seedance render from obviously-AI to professionally animated.
Steal forfinishing any AI-generated video clip before it ships
03:11concept
Reference-then-brief Claude prompt pattern
Drag a Pinterest style reference into Claude, then write a short brief describing what should happen in the animation; Claude asks clarifying questions before generating the final Seedance prompt.
Steal forany AI video or image prompt where you have a visual reference but not the words to describe it
How they asked for the click.
VERBAL ASK
20:17link
“If you want access to that Claude skill, those prompts, or if you want to access higgsfield.ai, go to those links in the description and you'll have everything that you need.”
Single end-of-video verbal CTA pointing to description links (Claude skill download + Higgsfield.ai referral link); no mid-video pitch.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
OTHER LINKSAlso linked in the description.
Visual structure at a glance.
