YT YouTube2Text

YouTube2Text → Transcripts → Comment faire du Motion Design PRO avec Claude Opus 5.5 (guide complet)

Comment faire du Motion Design PRO avec Claude Opus 5.5 (guide complet) — Transcript

by Maxence Tison · 4,382 words · 624 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00AI just killed motion designers. Help!
0:04Usually, I'm not a fan of clickbait
0:06hooks like that, but I have to say, the
0:08situation is dire for motion designers,
0:10because just look at what the latest
0:12Claude or ChatGPT models are capable of
0:14producing today. This type of video
0:17would have taken days of work, and thus
0:19would have cost thousands of euros. But
0:21from now on, you'll see that it's
0:22possible to recreate it in just a few
0:24clicks and a few prompts. And by the
0:26way, all the animations you're seeing
0:28around this video and the entire video
0:30edit were also done by my editor
0:31yesterday. And in this video, I'll show
0:34you step-by-step how I did it and how
0:36you, too, can install this system on
0:38your end. And I'll even explain how you
0:41can retrieve an entire database of
0:42animations ready to be reused with a
0:44single prompt. And incidentally, you'll
0:47also see that the argument which, until
0:48now, consisted of saying: "Yes, but
0:50once your video is ready, if you want
0:51to change things, if you have client
0:53feedback, it's impossible, it's too
0:54annoying," well, it's no longer valid
0:56at all today. Anyway, let's get started
0:58with the tutorial. So, buckle up. In
1:00this video, I'm going to show you
1:01different animation styles and three
1:03ways to do it. The first is to generate
1:05a video solely from a prompt. The
1:07second is to generate a video from
1:09images that we will animate. And the
1:11third method is to provide a reference
1:13video to draw inspiration from its
1:14style. Before diving into pure
1:16generation, you must understand that
1:18your AI, whether it's Claude or ChatGPT
1:21, isn't strictly capable of generating
1:23video. That, they don't know how to do.
1:24However, what they are very good at is
1:26creating code. It's probably even where
1:28they are the best. And in fact, today,
1:30there are tools that allow you to
1:32transform pieces of code into videos,
1:34into MP4 files. So there are two main
1:36tools on the market that allow you to
1:38do this. The first is Remotion, which
1:41is right here, which was historically
1:42the first, and the second is called
1:44Hyperframes, which is right there,
1:46which is a tool published by HeyGen,
1:47another tool we've already talked about
1:49together that allows you to make video
1:51avatars. So all you need to do is go
1:54and get the command that is either
1:56right here on the button on the
1:58Remotion site, or up here on the top
2:00right on Hyperframes. You copy it. You
2:03open a session in your browser on the
2:06Claude or ChatGPT side, you paste the
2:08command, and it will install it for you
2:10. In my case, it was already installed,
2:13so it just told me it updated since
2:15some things had been updated. Both are
2:17free and do essentially the same thing,
2:19but personally, I recommend you use
2:21Hyperframes, which is the one I will
2:23use in this video. And so, once your
2:25Claude or your ChatGPT has registered
2:27this system, well, that's it, it's able
2:29to generate video files for you
2:30directly from your conversation. By the
2:33way, these two tools, Remotion and
2:35Hyperframes, have another big advantage
2:37: they are completely free. It's not an
2:39extra subscription you need to take out
2:41to generate videos. It's just a little
2:44system that allows you to translate
2:45snippets of code into video. But,
2:47unlike other tools like X-Field,
2:49Magnific, Luma, or whatever you want,
2:50that allow you to generate images and
2:52videos, here, there is no generation
2:54system; it is directly integrated into
2:56your Claude or ChatGPT subscription,
2:58which will allow you to generate the
3:00videos we will see right after. And so,
3:03I suggest we do a first test for the
3:04first method, which simply consists of
3:06generating a video from a prompt. And
3:09so, as we are used to doing with AI, we
3:11will speak to them in natural language.
3:13Hello my dear Claude, I would like you
3:16to make me a motion design
3:18advertisement video for Santa Claus. So
3:21let's assume that Santa Claus is a
3:23business and I want you to make me an
3:26ad in an isometric 3D style. Uh, you'll
3:28see a little bit what that is. You can
3:31look for inspiration if you need to
3:33illustrate Santa's gift delivery system
3:36, where we see him flying over or going
3:39through houses, delivering his gifts
3:41through the chimney, and so on. The
3:45goal here is for you to create this
3:47motion design video, still in the
3:50isometric 3D style, to show us a bit of
3:52what you're capable of, as if this were
3:55a video for your portfolio meant to
3:58convince companies to hire your
4:00services for motion design. So
4:02obviously, I'm counting on you to make
4:05something impressive with zooms,
4:06illustrations, animations, and so on.
4:09Here, I'm using the Opus 3.5 model.
4:12Cloud, and for the effort level, I'm
4:14going to set it to the max to see a
4:15little bit of what it's capable of
4:17doing. A little later in the video,
4:20I'll show you some comparisons between
4:21Claude and ChatGPT, and between the
4:23different Claude models. Now, we just
4:26have to press enter to launch the
4:28generation. Before it starts working, I
4:34just added an instruction here to tell
4:36it to also create a voiceover, which is
4:38always nice in motion design videos.
4:41But I told it that to do so, instead of
4:43using its default models or what's in
4:45Hyperframes, which makes mediocre,
4:46quite robotic voices, it could use
4:48Google's Text-to-Speech models. I have
4:51direct access to it in my Claude
4:52because it's already connected via API.
4:54But otherwise, you can very well go
4:57directly to Google Cloud and ask it to
4:59create voices here. You have plenty of
5:01voices you can select from, you can
5:03provide the text to speak, and also
5:04give instructions here to customize the
5:06output regarding emotion, the tone of
5:08voice to use, and so on. So then, I
5:10launched it, and here is a little bit
5:12of what it created for me. No extra
5:13instructions were given in between. It
5:17gave me this result directly, which I
5:18suggest we watch together right now.
5:22Tonight, 2 billion children are waiting
5:24for a gift. A single delivery person, a
5:27single night. Impossible. Not for Santa
5:31and Co. At the North Pole, our
5:32workshops run all year long. Every gift
5:35is checked twice. Our sorting center
5:39knows every chimney on the planet. Our
5:41reindeer fleet is zero emission. 24
5:44time zones, one single route. And for
5:48the last mile, to the chimney,
5:50delivered before you wake up. Santa and
5:54Co., magic delivered on time.
6:01>> I don't know if you realize the amount
6:03of work it would have taken to churn
6:05out a fully animated 40-second video
6:08like this in an isometric 3D style,
6:10right? You're starting to understand
6:12what this is about. It's this style of
6:15video. It would have taken literally
6:17days and days of work. And here it is,
6:19just from a prompt without providing
6:21any references. Claude and the Opus 5
6:23model. Were able to make this video for
6:25us. It's just incredible. And if you
6:28need to provide feedback, well, you can
6:29use the interface we have right here in
6:31Editing OS, which is something I built.
6:33You'll also be able to access it under
6:36the video, allowing you to provide
6:37frame-by-frame feedback, a bit like on
6:39Frame.io. Let's say I want to point out
6:44that the banner here on the right isn't
6:47very visible, and I want it in black.
6:52Boom, I add it as a comment so that the
6:54next time it generates a new version,
6:55it knows that at the 10th second, at
6:57frame 304, there's something to modify
6:59in this video. Once you've provided all
7:02your frame-by-frame, second-by-second
7:04feedback, you can either click "copy
7:06for Claude" and paste it here, or click
7:08"send to editor" to trigger a new
7:10generation. It's just like you're
7:12talking to the editor you're working
7:14with. To get this interface here, which
7:17is Editing OS, you can also open it
7:19directly in your Google Chrome to see
7:22all your projects, provide feedback,
7:24and so on. I've already made a video
7:26entirely dedicated to this subject.
7:27I'll put it under the video for you as
7:29well. For the second video, we'll use a
7:32second technique which involves giving
7:34your AI reference images that it will
7:37animate and use to create its video.
7:41And for our second example here, rather
7:43than starting with Santa Claus, we'll
7:45start with a real app called Yuka,
7:47which you might already know, an app
7:50that lets you scan food and products to
7:52see if they're good for you or not. So
7:56that's what we're going to give to our
7:59friend Claude, and we'll feed it direct
8:01screenshots of the app so it can use
8:03them to animate and create a video for
8:06us. So what I'm going to do is copy all
8:09the images provided here so it has all
8:12the references. I'm also going to give
8:16it the website URL and tell it: "You
8:18are a motion designer," and again, I
8:21want you to create a video for me
8:23that's a motion design ad, an
8:25explanatory ad for what the Yuka app is
8:28, how it works, and what it's for. I've
8:33put screenshots right above from the
8:36App Store, which help to illustrate a
8:38little bit of what Yuka does. You can
8:42go look at their site, you can go look
8:43at the App Store, go try to understand
8:45what they do. It's an app that lets you
8:48scan all types of products to know if,
8:50well, they are good for your health,
8:51for the planet, and all that. And so I
8:55would like you to make me a 15 to 30-
8:58second video that is a promotional
9:00video for Yuka that will allow us to
9:03simply promote it. Don't rely solely on
9:06the images I gave you, they are more as
9:08a reference. I would like you to show,
9:12for example, when the app comes to scan
9:15a product, whether it's cosmetics or a
9:17product, a pack of cookies or whatever,
9:20and show us what happens in the app and
9:23so on. The goal here is for it to be an
9:26explanatory video about what the app is
9:28, how it works, what it's used for, who
9:30it's targeting, and so on and so forth.
9:32So it's your turn to play, my dear
9:34Claude. Likewise, we're going to set
9:36ourselves to Opus 5. We'll set it to
9:39max to give it all its full
9:41capabilities, and we just have to click
9:43enter. So, as you can see, it's
9:48starting to work well. Here, I gave it
9:51instructions again on how it could
9:52generate directly with Google's models
9:54rather than the default ones. And I'll
9:57show you once again what it created for
9:59me. Once again, on the first try. So we
10:01see that first it will start to create
10:03the first frames, the first images, and
10:05then go on to put them together. And
10:08here is the video that was created.
10:10>> What you eat, what you put on your skin
10:11, do you really know what's inside?
10:13With Yuka, you scan and you get a score
10:16out of 100 from green to red and you
10:18see why. Too much sugar, nine additives
10:20. Same for your cosmetics, risky
10:22ingredients are flagged. And if it's
10:25bad, Yuka suggests something better. No
10:26brand can influence the rating. Already
10:2985 million users. Download Yuka, it's
10:32free.
10:34>> Once again, the result from the first
10:36try is just stunning, without any
10:38feedback or revisions. What I found
10:40really well done in this video is that
10:42the voice is quite well synchronized
10:43with what's happening on screen. When
10:45it talks about the scanner, it's at the
10:47exact moment it uses the scanner
10:48function in the app, and so on. In
10:50short, this is the kind of thing that,
10:52once again, would have taken an entire
10:54team and whole days of work to achieve.
10:55And for the third video, we're going to
10:57do several things. First, we'll give it
10:59reference videos that it can draw
11:01inspiration from to recreate. And then,
11:04what we'll try to do is give exactly
11:06the same instructions, the same prompts
11:07, to different models, different levels
11:09on Claude, and also feed them to
11:11ChatGPT on Astra to see what it's
11:13capable of. So, for this third example,
11:16I suggest we take a very concrete,
11:18real-life case, something I really need
11:21: a video presenting Génération IA,
11:23which is my own training space for
11:25creatives who want to learn how to make
11:27ultra-realistic AI images and videos
11:29and use them in their business. So,
11:32let's get started with the prompt. Dear
11:35Claude, I need you to create a new
11:37motion design video to present my
11:40community, Génération IA. Given our
11:44conversations, our history, and
11:45everything you can find in my files,
11:46you already have quite a bit of
11:47information. What I recommend you do is
11:50also take a look at my website,
11:53specifically my community on Skool, to
11:55grab the images, the graphic DNA, the
11:58style, the fonts, the colors, etc. Also
12:00, understand the narrative—who it's
12:03for, what it is, what we do there, what
12:05should be highlighted, and so on, to
12:08create a motion design video aimed at
12:10presenting what we do. Tell who it's
12:14for, what the value proposition is, and
12:16so on and so forth. I'm going to
12:18provide you below with examples of
12:20previously created motion design videos
12:22that particularly inspire me and that I
12:25want you to use as inspiration to
12:27recreate your own video. So that's it
12:29for this prompt and for everything
12:31related to video references. I'm going
12:34to give it several videos that I simply
12:36found on Twitter, videos made with
12:38motion design that I thought were cool.
12:41So there was this one, and there's also
12:43this one for anything related to
12:44interfaces, very, very dynamic
12:45animation, and so on. There was also
12:48this one, which was a bit of a viral
12:50video on social media of Claude
12:52presenting itself as a motion designer.
12:55There was also this one, which I found
12:57quite cool, from a SaaS platform
12:59presenting itself. So I’m going to
13:02give it these four links directly and
13:03tell it to use them as inspiration. So
13:05there’s 1, 2, 3, and 4. Here we’re
13:09going to set it to Opus 3.5. Medium,
13:11and then we're going to generate the
13:13same video across different models. So
13:15we’ll launch that, and then I’ll
13:17simply copy all of this and open a new
13:20conversation. Here, we’re going to
13:22switch to Sonnet, which is the model
13:24that is supposed to be the best model
13:26from Claude. And likewise, we're going
13:28to go to, you know what? To Opus, which
13:30is supposed to be the best thing from
13:32Claude. And we’ll see a little bit
13:33what that gives us. And I'll do the
13:35same on ChatGPT, where I'll paste the
13:38exact same prompt, and I’ll give it
13:40the maximum capacity of GPT-4o, which
13:43is supposed to be the best model
13:45offered by ChatGPT. And all we have to
13:47do is launch the request. While it's
13:49working, what I recommend you do is go
13:51through your own exploration phase to
13:53give it some good inspiration. Good
13:56sources would be, for example,
13:57Pinterest, where you can type "Motion
13:59Design" plus certain keywords that
14:01match your style or what you like.
14:03You’ll find plenty of videos like
14:04this one or this one that you’ll find
14:07interesting, which you can then point
14:09ChatGPT or Claude toward so it can use
14:11them as inspiration for your own design
14:13. There are also other sites starting
14:14to emerge, like this one. I’ll put
14:17all the links just below, which lists
14:19plenty of animated-style videos. Here,
14:23we have an explainer video on how to
14:25make such-and-such cocktail, and so on,
14:27or there’s also this site called
14:29whatships.com, which is quite
14:31interesting and actually gathers
14:33startup launch videos that look like
14:35this, which are typically the type of
14:37people who need to make motion design
14:39videos. Well there you go, these are
14:42things that can inspire you and once
14:44again things you can steer your cloud
14:45towards so that it comes and creates
14:47videos for you. So here again, it
14:49starts doing its research work. It went
14:52and opened my community's AI generation
14:54site in its browser and also asked me a
14:57few questions about the format,
14:59duration, scripting, the voice it
15:01wanted to use, and so on. And it got to
15:04work. It offered me different
15:06voiceovers that I was able to choose
15:08from. It proposed a script that I
15:09validated, and I told it that it could
15:11go ahead and start the generation. And
15:13if I scroll a little bit, you'll see
15:15that it created this video for me. AI
15:18can already create images and videos
15:20that look real. Hardly anyone knows how
15:23to use it. Génération IA is the
15:25community that teaches you how. Ads,
15:27UGC, product visuals, over 50 lessons
15:30from beginner to pro. One live session
15:32per week, ready-to-use prompts and
15:34workflows. Then we connect you to
15:36brands that pay AI creators. Over 200
15:39creators. Rated 5/5. Génération IA,
15:42try it for 7 days for free. Really not
15:48bad at all. It's the kind of thing I
15:50could run as an ad on YouTube or
15:52elsewhere. And besides, I even asked it
15:54to make a short ad version for the free
15:56masterclasses I organize. I'll show you
15:58the result. So, you’ve realized that
16:00AI can make ultra-realistic images and
16:02videos. Yet, when you try it yourself,
16:04the result is really not great. The
16:06problem isn't the tool, it's the method
16:08. And this method, I'm going to show it
16:10to you in a live session where I'll
16:12share my screen, my tools, and my
16:13entire process step by step. Over 2,500
16:15people have already attended one of
16:17these masterclasses and it’s 100%free
16:19. You just have to click right below
16:21this video to book your spot and I’ll
16:22see you soon.
16:23>> Anyway, honestly very impressive. It is
16:25literally an ad that I am going to use
16:27in my Meta account to drive people to
16:29my free masterclasses. To test this and
16:31take it a step further, I asked it to
16:33make me an ad, once again on the fly,
16:35for this tool Flora that I use a lot,
16:36which is a tool that lets you do this
16:38kind of thing, connecting many AI
16:40models together, having a nodal
16:41function, so making them work with each
16:43other, and so on. And here is exactly
16:46the video it made for me. Honestly,
17:12very impressive. And besides, what's
17:13pretty crazy is that it went to pull
17:15from my own account on Flora. These are
17:17real images, real creations that I made
17:19, that I created myself to illustrate
17:20exactly what is possible to do with
17:22this tool. Now, let's see what Claude
17:24did, but with the Fable 5.1 model in
17:26the ultra version. So, it went through
17:29a bit of the same process; it started
17:30by proposing voices to me, it wanted me
17:32to approve the script, and so on. And
17:34here is the delivered video.
17:35>> Did you see what it can do with images
17:37and video? And when you try it yourself
17:39, it shows. I'm Maxence, ex-Google and
17:42TikTok with 600,000 followers.
17:44Génération IA is my community. It
17:46teaches you how to create AI images and
17:48videos that look real. You learn how to
17:49turn it into income for creators,
17:51freelancers, and entrepreneurs who want
17:52to make a living from it, whether
17:53you're a beginner or already a pro.
17:54Inside, there are over 50 lessons in
17:56seven modules. One live session per
17:58week with me, a prompt bank, end-to-end
18:00workflows, and personalized feedback on
18:02your creations. As a bonus, Editing OS,
18:04my AI video editor. $ 5,000 in credits
18:07for your tools and brand missions
18:08looking for an AI creator. Over 250
18:11creators are already sharing their
18:13results. 5/5 rating. Companies pay me
18:16€ 3,000 for this. You pay $ 87 per
18:18month. Price locked in, 7 days are free
18:20, and you can quit in two clicks. Head
18:23over to school.com/generationia. I
18:30don't know what you think, it's not bad
18:31, but personally, I find it less good
18:33than the previous version. Well, for
18:34one thing, it's a longer video. It
18:36lasts 53 seconds, so it made a bit more
18:38effort to create something more
18:39comprehensive. Uh on the other hand,
18:41well, it made an effort in terms of the
18:43animations to include, especially these
18:45kinds of little animations. We see the
18:47little switch turning on and so on. But
18:50personally, I notice there are still
18:52more design errors in this Fable model,
18:54paradoxically, even though it's
18:55supposed to be better than Opus, which,
18:57uh, is visibly a bit, uh, more advanced
18:59in terms of motion design. Or well here
19:01, for example, it's a perfect
19:03illustration of that, where we have the
19:05black text "free trial" that is on top
19:07of the text above it which has a black
19:09background; so, which makes it look
19:11like the text is a little bit cut off.
19:13Here is another example, for some
19:15reason I don't know, it put the text "
19:16personalized feedback" which is
19:17completely cut off on the right, which
19:18isn't whole. modules
19:20>> or even there, at that moment, the text
19:22"7 modules" which arrives on top of the
19:24module images, which is a little bit
19:26weird. Anyway, it seems like it put a
19:28huge amount of effort into animating it
19:30in every direction and it puts in lots
19:31of things, but the result I find is a
19:33little bit worse than Opus 3.5. And on
19:35top of that, it consumed more tokens.
19:37>> Your ideas, your images, your videos
19:40created with AI. For real, Génération
19:44IA, the community of creators,
19:46entrepreneurs, and brands. From the
19:49first prompt to the final edit, take
19:51action. 50 lessons, live sessions, and
19:54feedback to progress together. For your
19:57clients, for your brand, your next
19:59creation starts here. Join Génération
20:04IA, 7 days to try it out.
20:07>> Well, that was the ChatGPT version. I
20:09don't know what you think honestly,
20:11it's not, it's not incredible. It's not
20:13that great, I find, compared to what we
20:15just saw on the Claude side, and
20:16especially on Opus 3.5. I don't know if
20:20it's because the ChatGPT models are a
20:22little less made for this and are less
20:24good at it, or if it's also because I
20:26use Claude much more on a daily basis
20:28and so it has much more context on what
20:30I do, what I like, etc. But for now, I
20:32would say the big winner of this battle
20:35is clearly Opus 3.5. So far. The video
20:41for Santa Claus, the video for Yuka,
20:43the video for uh Génération IA, or
20:45even the video on Flora that I made as
20:47a bonus. It's clearly quite impressive
20:50what is possible to do today. And just
20:52to show you what this means on the
20:55market today: on Malt, which is a
20:57freelance platform, freelance motion
20:59designers generally charge around €
21:01400 per day (from € 350 to € 450
21:04for the most experienced) to make the
21:06exact type of video you are watching
21:08right now. Keeping in mind that a video
21:12will generally take them at least one
21:13day for a short video, or even several
21:15days of work. I can also show you some
21:17examples of what people usually charge
21:19for this type of video. Here, I am on
21:22Fiverr, which is another freelance
21:24platform where we have someone offering
21:27to make exactly the type of video you
21:29are seeing here, starting at € 748.
21:32And it goes up to € 1,500 or even €
21:342,000 per video. Here as well, the base
21:37format is € 1,100, then € 2,300,
21:39then € 3,000 per video, again for
21:41this kind of content, which is exactly
21:43the type of video you have just learned
21:45how to make. And to give you one last
21:49example, we have agencies here, like
21:51Agence 1600, which is a fairly
21:53well-known agency in France that really
21:56specializes in Motion Design, and they
21:58are used to selling this type of video
22:01to startups that want to launch a new
22:03product, announce a new feature, etc.,
22:06for € 4,500 for a 60-second video, as
22:08you can see here, or € 6,000 for a 60
22:11-second video with a few extra bonuses.
22:15So, this is to give you a bit of an
22:17idea of how much what we just looked at
22:19together regarding motion design
22:20generation with AI is revolutionizing
22:22this job and this industry. And to show
22:26you how much of an opportunity there is
22:28for all the brands that couldn't afford
22:29motion designers until today, whom you
22:31will be able to approach and say: "Well
22:32, there you go, from now on you can do
22:34motion design because it will cost less
22:35and I will do it for you." Or for all
22:39the brands that are already making this
22:41type of content, whom you will be able
22:42to approach and say, "Well, from now on
22:44you will be able to make it for less,
22:46or I can do it for the same price but
22:47in a much, much more advanced way." If
22:50you want to get my system that I call
22:52Editing OS, which will allow you to
22:54manage all your video creation projects
22:56and access lots of animation libraries
22:58and so on, you just have to join
23:00Génération IA. It's free for 7 days.
23:02You can come and watch everything for
23:04free, the training videos, get the
23:05workflows, the tutorials, and even come
23:07to the live sessions I host every week
23:09to answer your questions and so on. If
23:10you like it, you stay, that's great. If
23:12you don't like it, you just leave.
23:13After 7 days, you won't be charged
23:14anything at all and we'll remain good
23:16friends. I've put the link in the
23:17description and I hope to see you very
23:18soon.

About this transcript

This page contains the full transcript of Comment faire du Motion Design PRO avec Claude Opus 5.5 (guide complet) by Maxence Tison, generated from the public captions YouTube serves with the video. The transcript has 4,382 words across 624 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.