S2. Ep41 - How to Create On-Brand Images and Videos With Claude (No Designer Needed)

Katie (00:27)

Hello and welcome back to another episode. Hi, hello, I'm Katie, and as always I have Noel here with me. Hi Noel, how are you doing this week?

Noel (00:36)

I'm very excited this week. How are you?

Katie (00:40)

Yeah, so there's a big reason why Noel is excited this week. Something has just happened in the AI and automation world. He's been glued to his screen, I wasn't allowed to interrupt, and he's going to share that excitement with us. Do you just want to start with that?

Noel (01:07)

I think we should, yeah.

Katie (01:09)

Can you contain your excitement and not start with that? I think that's really the question.

Noel (01:17)

So I guess in terms of other relevant news, there's not really been a lot happening. But there is one thing. I can hold on for one item and then we'll get into it.

Katie (01:28)

Okay, wow.

Noel (01:30)

So yesterday I opened up Claude and I had a little pop-up message that said, claim your first $100 in API credits. And I was like, what's this about? What's happening? So if you have a Claude Max or Team plan, basically they're giving away money for you to spend on their API.

Noel (02:00)

The base is $100, which is very, very generous. I would never get through that in API usage. But if you have the highest Max plan, that goes up to $200. And then Teams can go up to $500. So they're encouraging you to use their API for your apps and things like that. I thought it was a really nice little gift to get this week.

Katie (02:27)

Do we know if everyone's getting that?

Noel (02:30)

It's only if you've got the Max or Team plan at the moment.

Katie (02:34)

Okay. So anyone who's got the Max or Team plan is getting a hundred dollars worth of credit.

Noel (02:42)

As a minimum, yeah. Well, Teams is a little different. If you have a Team standard seat, it's $20 per seat, and a premium seat is a hundred dollars per seat, up to a maximum of $500. Either way, you're getting, I would say, a substantial amount of API usage credits just gifted, which is very nice. I won't say no.

Katie (03:11)

Okay.

Noel (03:13)

But I guess we'd better get on to what I'm really excited about, if that wasn't enough.

Katie (03:18)

Yeah. Free credits, but that's not what you were very excited about.

Noel (03:25)

No, definitely not.

Katie (03:26)

Okay, so today was the Make Waves conference. That's the annual conference by Make.com. We didn't actually attend this year, it was in Prague. A few things meant that we weren't able to attend this year, unfortunately, and go and see our friends at Make. However, they do livestream the first part of the conference. It's about an hour and a half, where the CEO comes and has a chat, and some of the other... is it directors?

Noel (04:10)

The product managers for different areas.

Katie (04:14)

Yeah. So they come and tell you all about their really big updates, what Make have got planned, usually for the next year, and showcase the big things that are coming to Make. Last year they showed us Maia, which was their agent where you could speak to the agent rather than having to build it yourself, which we were all like, oh my god, this is amazing.

Noel (04:52)

Yeah, it was really good. It is really good as well. You can use it today.

Katie (04:57)

So that's what Noel's been excited about. He's been glued to his screen watching that today. So Noel's going to tell us what Make announced and what you need to know as a business.

Noel (05:15)

Yeah, lots and lots of things were discussed. Lots of things that I'm excited about, and other things that are just, oh, that's nice. They were almost like footnotes. I think in previous years some of those things would have been massive, whereas this year there were some very big topics that they covered.

Katie (05:38)

Do you think that's because technology, AI and automations have moved on so much in the last year? When I think back to what they were saying to us last October at Make Waves in Munich, about when this is going to happen and when they'll release that. Do you think it's just moved on so much that some things we would have considered big a year ago are now, like you're saying, footnotes, because there are so many other bigger and more awesome things?

Noel (06:25)

Yeah. So I think probably the biggest thing, and the thing they're worried about the most, is losing their customer base. If you think about it, these days with Claude, ChatGPT and Gemini, you can connect your apps in and use it as an agent. You can automate quite a lot of things in your daily life and work life using those sorts of tools.

Noel (06:50)

So they're looking at it from the point of view of, how do we retain, or look to increase, our customer base in the future? Because as we know, AI changes every single week. They can't really keep up with that, but they need to restructure in a way that actually makes sense for people to think, actually, yeah, I think we should stay with Make for the long term, instead of just using ChatGPT or Claude, with Cowork or Work.

Noel (07:20)

So the big thing this year was transparency. They were basically showing that AI and automations currently are kind of a black box in your business. It does stuff in the background, and it either works great, or sometimes it might not work that well, but you don't really get to see that much of it.

Noel (07:45)

So the big thing they're pushing is the AI glass box, and they pushed that throughout the entire keynote. They're giving you more visibility and transparency into how not only the agents in Make are working, but also your automations as well, so you can make better business decisions and make your life more efficient, I guess.

Noel (08:10)

Which is really quite interesting, because we can all tap into Claude and ChatGPT, but who really clicks on the button to find out exactly what its thought process was and what happened? What did it actually do in your inbox if it went off to Gmail to go and find something? You don't really get to see that very well, and they're really opening all of that up. So it makes it easier for businesses to get on board with Make. That was really quite interesting to hear them talk about.

Noel (08:50)

But what I would say is there's a big push for agents. They're not really that pushy on building automations. And they did make a really interesting point, where they said that in the future, drag and drop builders will be gone. We won't be using Make how we do today. It will all be done through a chat interface.

Noel (09:21)

And I think the introduction of Maia was step one. That would automatically build your automation, connect your accounts for you and guide you through that process. So that's really, really quite interesting.

Katie (09:33)

Yeah, because if you remember what they were saying last year at Make Waves, they were asking people for feedback, which is what I love about Make. They do always ask for people's feedback and they do implement those changes. And they were saying that a lot of people were struggling to use it, or not using it as much, because of the tech restrictions. They didn't have the knowledge to be able to do the drag and drop and build an agent. And even now, that to me feels like such an old fashioned way of creating an agent. Does it to you as well?

Noel (10:23)

Yeah, it does still seem a bit clunky, doesn't it, building it manually?

Katie (10:28)

So clunky. I'm not surprised that that's going to be phased out, because no one wants to be using an interface like that, when we can speak into something and it can create an agent that way instead.

Noel (10:55)

Exactly. And there's a bigger release that goes a bit further on that, but we'll come on to that one closer to the end, because it's quite a big thing.

Noel (11:06)

But one of the other things they wanted to do, as well as the transparency part of it, was help you make better decisions with your AI and automations, your agents and all that sort of stuff. So they brought out a feature called Process Intelligence. They didn't actually show it working live, so I don't think we'll see this for probably a year. It's going to be a while.

Noel (11:42)

But essentially what they were saying you could do is record part of your manual process on your laptop or whatever, so it records your screen. It'll watch you manually moving data from one application to another and all that sort of stuff. It would then take that information and also look at the current scenarios you have built into your process. And then it would highlight those areas and go, right, what we can do is, instead of you doing all this stuff manually for an hour, we could create this automation that will take that part of the job away from you and make it more efficient, so you've got more time to do something else.

Noel (12:30)

I thought that was really, really quite interesting to see, and I'd love to see how it would actually fit in and work. Because in Make at the minute, you kind of build processes that are outside of your company processes, if that makes sense. They help you facilitate it, but there's no way of bringing the two things together.

Noel (12:59)

So I think they need to bring that in as well. You could say, this is our overall flow for, I don't know, onboarding a new lead. That could be documented in your company, and then you link that in with Make and have this sort of Process Intelligence making it better and more integrated. I thought that was really quite interesting. I look forward to trying that one out.

Noel (13:31)

There are also lots of things if you run a team within Make. You might have a couple of people in there who all have different roles. At the minute it's really difficult to restrict what they have access to. Either they have access or they don't. It's on or off when it comes to the apps and things like that.

Noel (13:55)

So what they're putting in place is a way where, let's say you've got somebody that works in HR, they can only access the apps and the agents that are relevant to that particular job role. So they can't go off and run a sales agent from an HR role, which makes complete sense.

Katie (14:16)

Yeah, that does. Not everyone needs access to everything.

Noel (14:20)

Exactly. And they can build stuff based on their specific role. So HR might think, actually, we could do with an agent that does this particular function, and they build it out, and then that would only get shown to that team or that individual. I thought that was really quite clever. It's one of those things you just wish was there from day one, don't you? But I guess they've had a lot of enterprise and big teams feeding that sort of thing back to them. So it's really interesting.

Noel (14:58)

But there are lots of other little updates coming out as well.

Katie (15:03)

Okay. Should we quickly run through some of those?

Noel (15:07)

Absolutely. There aren't that many, to be fair. But they are bringing back human in the loop from two years ago. At the first Make conference I went to, they talked about human in the loop, and the way it worked was you could add it in there as a check function. It would produce a URL that you could send to Slack or your email or whatever. Then you'd click the link and it would open a web page hosted by Make, and you could amend it, approve, deny, all that sort of stuff. And that would trigger the rest of the automation. Which, fair enough, that works. But that then got hidden in the enterprise layer and didn't actually come down to anyone else on a Team plan or anything like that.

Noel (16:00)

With this new one, they're letting you integrate it into things like Gmail and Slack directly. It just pops up with an approval notice, you click approved, done, and the automation continues. So there's no need to navigate to a random website anymore. It's there in the apps you currently use, which I thought was awesome. And obviously, based on whatever you give back, you could then route it to do certain things in the rest of the automation.

Noel (16:30)

The other one, which I thought was really cool, and I sort of wish it had been in there since day one, is this. When you run an automation and you're trying to test it out, you lose the information. So if you process a row in your spreadsheet or database and think, I'd like to rerun that, it's actually quite hard to do. You might need to manually go back and change certain things in that database or spreadsheet to allow Make to find it a second time.

Noel (17:04)

So what they're doing is bringing out data pinning. You could say, great, that run was awesome, but I want to test it again. So you pin the data and run it again. And it will either run the entire automation, or you could set a point and go, only run it up to here. Don't go through the entire process flow, just do the first three modules. And it will just do that for you.

Noel (17:30)

I thought that was going to be really quite cool for testing things out, because the number of times I've had to manually go back and change data in a database just to see it run again... it was so frustrating.

Noel (17:50)

They're also talking about better knowledge bases as well. They didn't show that in any way, but they did discuss it, where you could have knowledge bases for certain functions in the business, and the AI agents you create within those roles and areas of the business would have access to that information. So they're making a big move on that as well, which is pretty awesome. That was really the bulk of the new things that are coming.

Noel (18:27)

But there was one other, which they left to the end, of course they did. This was called Make OS, which got lots of applause. People got excited for this one.

Katie (18:40)

Okay.

Noel (18:42)

So essentially this is an app that you would download onto your computer. I don't think you would actually use it within Make. You'd download it and it would basically look like ChatGPT or Claude does when you download those. There'd be a chat screen, and you could have a conversation with your scenarios.

Noel (19:05)

The example they showed was, could you get me the information for the last quarter's sales? And it went off and gathered all the information from the current scenarios and delivered that back. Then they showed that the next time, they asked the exact same question, and Make OS popped up with a little box saying, you've asked this twice now on a regular basis, so why don't we create an agent that will help you do this, and maybe create a scenario which will run every month and get you that information. So it was really proactive in what it was doing, which I thought was really cool to see.

Noel (19:50)

The other thing they showed was a process. It didn't look like a Make.com automation at all, but it was highlighting things within your automation, saying, if you did this, you could save 50% on your AI usage within this particular workflow. And I was like, that's quite clever. I'd love to see how that works in real life.

Noel (20:27)

But it did just look like a normal flowchart, which I thought was a bit boring for Make. I was expecting it to be a bit more wow, a bit more visual. But who knows, it might change. I'm sure it will.

Noel (20:41)

There was lots of transparency built into those processes in Make OS as well, so you got to see everything that was being done. And if you have a team, you could share the agents that you've either created, or that have been suggested to you to create, with your teams as well. They could go off and chat to them, and you give them little names and avatars. The sales agent was called Penny, with a little cartoon character.

Noel (21:21)

It'll be interesting to see what that turns into. But I think this is the point where they're moving away from the visual builder, and you would actually use Make OS to build and manage your automations. So lots in there.

Katie (21:41)

Really interesting, though. What are you most excited about?

Noel (21:51)

I think Make OS, definitely. I'd love to get my hands on that. And maybe also Process Intelligence. I'd love to see how that works.

Noel (22:04)

With all of this, you can sign up to the waitlist to get access to the features we've talked about. We'll pop a link to that in the show notes, so if you're a Make user already, you can jump in, put your name down and hopefully get beta access soon. Usually they give some of them out quite early. It depends on how far developed they are.

Katie (22:29)

Other times you're waiting quite a long time, aren't you, if I remember correctly?

Noel (22:34)

Yeah. Some will come out January, February, and then the rest is usually in the summer, I usually find. Exciting times.

Katie (22:40)

Very exciting times. I think that might have been one of our longest update sections ever, so if you're still with us, thanks very much.

But now we're going to move on to the main part of this podcast episode. So Noel, do you want to explain a little bit about what we're going to be talking about?

Noel (23:18)

So what we're going to go through is how you can create consistent images, videos and animations using Claude. You might not have thought that was possible, because it's text-based and code, but it is 100% possible. I've been doing it for about two months now, so I'm a bit ahead of the curve, I think. Hopefully.

Katie (23:42)

Toot your own horn.

Noel (23:44)

I know. Well, someone has to, as I always say.

Katie (23:51)

Okay, so what can it do? What do we need to know about it? Tell us.

Noel (23:58)

So I guess let's start with the branding first. I think that's probably the most important part to get right.

Katie (24:06)

Before you start, is this available to anyone on any plan?

Noel (24:14)

Great question. I think for the design system you have to be on a paid plan. However, creating the videos and images manually on a free plan should be very much doable. It's just asking Claude to do it, there's nothing gatekept. I just think the design system is behind a paywall.

Katie (24:39)

Okay.

Noel (24:40)

So to get started, let's go into the design system first. Obviously, when we're creating images and videos for our business, we want them to be consistent in form and style. And they've created a new way of doing this.

Normally, what I would do is say to Claude, look, here's my website, can you create me an HTML document which details all of the brand elements from that website? It will go through, and on our Make Automations site, for example, we've got Post-it notes and pens and pencils and little objects and assets around. So it would put all of that into the HTML document in code, so every time we want to use it, it looks exactly the same, which is just what we want.

Katie (25:38)

Yeah, because I find that can be really difficult with images and photos, even if they're your own or even from a stock website. I feel like it can be really difficult to tie a particular look, or even brand colour, all together.

Noel (26:03)

Yes, I've definitely had many troubles trying to do that in the past. I'm not very good at this sort of stuff normally, but thankfully Claude can help me out.

Katie (26:15)

No, I think as well, your brain works in a very logical way. Even when I've said to you about something at home, like, this is the vision I have for this room, you just cannot see it at all. And when we've gone to look at houses and I'm like, oh gosh, this house has got so much potential, we could do so much to it, you cannot see that at all. That part of your brain is not activated.

Noel (26:54)

But the engineering process part works. It's like, actually, that room is going to take ages to redo, there are all these different things that need doing. So I'm good at that part.

Katie (27:04)

Yeah, whereas I don't think about that at all. It's just, this room would look great if we did this to it and that to it and added this, this and this. So those of you who are listening can probably guess who does the decorating in the house, and says, right, we need to redecorate this.

Noel (27:30)

Hmm, it's not me.

Katie (27:32)

No, it's not you. So it totally makes sense that someone like you struggles with putting a cohesive brand together. Whereas for me, I think that's brilliant and I get really excited about it. So this tool is really handy if you're more of a logical thinker, like Noel.

Noel (28:06)

Yeah, definitely. But they've kind of hidden where they've put this design system. So one way of getting to it: if you're in Claude and you hit the plus button in the chat window, you'll see an option for design system. In there, you can create a new one using Claude or Claude Code, or you can manage your existing ones, or select one if you've already created it.

Noel (28:38)

But let's say we're starting afresh. You'd create a new system using Claude, and it loads up a new chat window. It gives you the option to type away and tell it all the different things about your business, the colours, the fonts. If you've already got documents, maybe you've created that HTML file, add that in there. My Make Automations HTML file is absolutely gigantic, so I threw that in there and thought, right, this is going to stretch it, can it do it? And it did.

Noel (29:16)

There are loads of different things you can add into it. You can link folders and all sorts, so you can come with absolutely everything. And what it does is create this design system almost as a standalone document, so you can scroll through it.

Noel (29:35)

If you give it access to something like your web address, it will go off and find patterns in how things are laid out on the website. It'll be like, they usually do this in this way, or this message is always highlighted in this text first and then the subtext. It's picked up on loads of different nuances. In the end, it will come up with all of your colours, typography, spacing, shadows, components, image assets. There's tons and tons of stuff it comes out with, and it's really comprehensively built.

Noel (30:10)

But where it comes into its own is when you start a new chat, hit the plus button and select your design system. You'll see it pop up, there's a little palette icon with the name of your design system. And then you can just say to Claude, can you go off and create me an image for this particular blog post? And it will use that design system to do it.

Noel (30:45)

So you're not having to reinvent the wheel and go, where's that HTML file, I need to pop that in there again just to make sure it's got the latest version. With these design systems, it's all nicely interlinked, and you just have to remember to tick a box. It's all done for you.

Katie (31:07)

Great.

Noel (31:08)

I thought it was incredible when I got that working. I was like, where has this been all my life?

Noel (31:20)

So I guess next is, how can you create an image or a video in Claude? It's quite straightforward when you start thinking about it. All you've got to do is ask. Could you create me an image? It could be for an article, it could be for a flyer, or a poster you're creating, anything you wish. And it would use that design system to start creating that image for you.

Noel (31:50)

Like I say, Claude isn't an image or video platform. What it does is build that image in code first. It'll show it to you almost like it's built it as a website, like an HTML file. And you can then look at it and go, actually, I didn't want this element, or this needs moving further left or right, little minor adjustments.

Noel (32:18)

And where it stands apart from AI imagery, creating it through ChatGPT images or whatever, is that if you say to Claude, just move that over to the left a bit, it just goes into the code, moves it, job done, everyone's happy. Whereas if you did that through an AI image model, it would have to completely redo the image. It gets painful.

Katie (32:42)

Oh god, yeah. The number of times I've just thought, I want some imagery, and for the life of me I cannot find it anywhere on the internet. And my last resort has been something like ChatGPT, and I've explained it in so much detail, and then it's loaded it and I'm just like, what the... And you've just asked for one change, and it changes the whole blinking thing, and I'm like, no.

Noel (33:24)

Yeah, or it starts adding extra fingers on someone's hand.

Katie (33:24)

Or it's not even that. It could just be adding different things, and it's like, no, no, no.

Noel (33:34)

It's horrendous. At least with this method in Claude, you can get those minute tweaks in and not have to worry about it.

But you can also add your own images as well. Maybe you want to add your headshot to a particular thing. You can just add that in and say, put that in there, in whatever corner, and it will go off and do it for you. And then it will also output it in an image format, so you can download it straight away and put it wherever you want. And I guess if you've got the connectors set up in your Claude account, you could get it to post it somewhere, those sorts of things.

Katie (34:21)

What are your thoughts on how people are actually going to use this? Are people going to be taking other people's photos? Where's the line between people using this genuinely for their own business, and people taking other people's photos, say their listings or content, and then...

Noel (34:51)

Rehashing it.

Katie (34:52)

Rehashing it. Copying it without permission. All those things.

Noel (35:01)

Yeah. I guess those problems have always existed.

Katie (35:09)

Yeah. But I think this just makes it easier for things to be stolen, including your own image. I know a lot of people on Instagram have fake videos that are supposedly of them, but they're not. Someone's downloaded all of their Instagram videos or TikToks, and somehow they've managed to create a new one, and it's selling something really random.

Noel (35:45)

Yeah, I think you were selling Bitcoin once, weren't you, on one of your accounts?

Katie (35:49)

Yes. But it's scary, because loads of those popped up. They were just using my photo, so I feel like it was pretty obvious to everyone that it was fake. Whereas I've seen actual videos being recreated and posted as that person. And I was wondering, do you think this is going to increase that sort of behaviour?

Noel (36:23)

Yes. I think with any of the normal AI models for image and video, that's definitely getting worse, isn't it?

Katie (36:33)

Yeah. And then obviously you're saying you could put photos of yourself into Claude, but then AI has photos of you. How are you protected with that?

Noel (36:52)

Well, there are always the protections in your account where your stuff isn't used to train future models, that's always in there. So no one's really stealing that information from you. And for myself and most other business owners, most of this stuff is already out there anyway. My headshot's already out there. If someone wants to use it, they're welcome to it.

Noel (37:20)

I just think, if it's already publicly available on my website, I don't really have much of an issue with AI going off and using that asset and remixing it in its own way. For me, I don't have much of a problem with that, but I get that some people would, and wouldn't like that.

Katie (37:43)

Yeah, I mean, I would.

Noel (37:46)

Yeah, definitely.

But the way Claude's doing it is, it's just building it out as if it was your website, and then creating it as an image asset.

When it comes to video, that's a little bit different. They're bringing out something called Motion in Claude. It's currently only available on Team and Enterprise accounts, but I'm sure it's going to come to everyone else eventually.

Katie (38:19)

Okay.

Noel (38:20)

Essentially, it allows you to create a storyboard for animated content. Say you were at an event and you've captured loads of B-roll from it. You could put that into Claude and say, at this point I want to add in an animation that shows one of the key talking points or key topics, maybe a QR code at the end, things like that. You add all of that in and create it the same way you would an image. Say, I want to create a video, this is what I want it to tell, here's the video asset, and it will go off and do it. And just the same as with the images, the edits are really easy, because they're in code, so it can figure them out really quickly.

Katie (39:16)

Okay, so go and check it out, and let us know how you get on if you do. Is this going to be a lifesaver for you? Are you someone who hates putting branding together, or hates using Canva, or maybe just lacks those skills? Do let us know. Send us an email, hello@makeautomations.ai, or come over to our free LinkedIn group, which is called AI Automations for Business. Everyone is welcome.

But we really hope you've enjoyed this week's episode, and we will catch you next time.