How to make AI mockups that don't look generic in FLORA

Lightly edited from the recording for readability. Wording is the speakers’ own.

Hi, everyone. Welcome. Welcome to our second AIGA New York FLORA Lab with Catherine Chung.

I'm Olivia, leading education here at FLORA, and we are so excited to have you all tonight, from all over the world, to get started on the FLORA canvas. Before we dive in tonight, I want to say once again a huge thank you to AIGA New York for partnering with us at FLORA, for producing this series of free, hands-on FLORA Labs, free to the public usage, given during the session. AIGA, as many of us know, has been home to the design community for decades and we really do share the same goal with AIGA at FLORA, which is to make sure that designers stay at the center of how creative work gets done as these tools evolve. And so we both really believe that starts with both product access but also education, which is what brings us here tonight.

So over the next five weeks, every Tuesday from six to eight, we will get our hands on the canvas with a different part of the creative process led by Catherine and Sasha, our forward deployed creatives here at FLORA. Go ahead, Kat. Awesome. Thanks, Olivia. Hi again.

Hi everyone. Good to see you either brand new or returning from last workshop, but yeah, today's going to be mockups are like one of those like love-hate type of things. As designers, We hate to look for the perfect one and we hate to kind of like mess around with smart objects and Photoshop. So thankfully, mockups, I feel, have gotten a lot easier with AI. So definitely wanted to make sure that we had a dedicated workshop just revolving around this stuff, especially because design case studies rely heavily on mockups.

Presenting design concepts to clients relies heavily on good, high quality mockups as well, and nothing kills a good design than a really bad mockup. So hopefully, after today you're a little bit, you have what you need to kind of upgrade that mockup game and all that sort of stuff. So, yeah, basically exactly what Olivia said. I'm going to walk you through a few things, but then we're going to go into the canvas together. So give me around like 15 to 20 minutes of some intro spiel and then we'll get right into it.

So basically, welcome to our second workshop in partnership with AIGA New York. Again about our partnership. FLORA is proud to partner with AIGA New York and provide designers across the country with free resources to feel ready for what's next in their field. These workshops are free to register, free usage for every class, no AI or FLORA experience is needed, so you can come in at any point and just kind of learn something new. It's also led by creatives for creatives, and no models on FLORA train on your inputs and outputs.

So that's a really important one. You can feel good about using FLORA today. None of our models train on anything that you put on the canvas or that you generate on it. So quick rundown of today's agenda. A little bit about us.

Why nodes and preparing to use them, exploring the canvas and the workspace and creating an editing mockups. So I, if you're returning, you might like a few of these things kind of might seem familiar, but I try to change things up a little bit every class just so I keep it a little bit fresh, but the same message kind of translates. So stick with me here if you've kind of heard of a few of these things before. But hopefully you kind of hopefully, the more you hear it, the better you feel about understanding how to use FLORA. So there's also that.

But a little bit about us if you haven't heard of us before. FLORA is a node based creative tool that brings the top creative AI models across text, image, video, audio and even 3D into one intuitive interface to accelerate creative production without compromising on craft. So FLORA has a few entry points. We've got the canvas, which is a node based infinite style canvas that allows you to visually program repeatable workflows, freely control how media interact or generate, and allows for a more hands on experience. We also have MCP connection using, which is where you use a third party LLM application to use FLORA to execute workflows and generate images with existing context already built out.

We also have create slash generate, which is a chat interface that allows you to generate images using any model we have on the platform. And then there are more to come. So, as the AI world kind of evolves, we're constantly trying to bring the most upfrontand then latest and greatest type of interfaces to our users. So definitely stick around. In our near future, we're going to be dropping product releases left and right, so we'll have- we'll be having more of these workshops revolving around new product drops and all that stuff, so you'll have a reason to pop in all the time.

But with that all being said, today we're going to focus on the canvas, because understanding the basic function of the nodes on our canvas helps you get the most out of the tool when you explore agentic workflows later on. So a few things to understand about AI and mockups. So AI models are amazing at photorealistic setups and compositions, but the results can be generic. The methods I showed today will help you fight off the typical AI look to mockups through prompting model selection and using meeting full of references. So an example of this is the image on the left was just generated after one run with Nano Banana Pro, and you look at it and it's pretty flat, Not that not that great looking.

But then on the right I use same prompt, a different model and a little bit of color correctionand it all of a sudden kind of it had a little bit of a little bit more to it. Another thing: sometimes not using the word mockup gives you exactly what you need. So the point of a mockup is to create an accurate depiction of an object or design in context to an environment. AI models might get caught up in the word mockup and it'll give you something generic. So a good example of this is on the left.

I wrote a prompt, something: give me a mockup of a cardboard box, blank whatever- And it gave me a mockup It was. I feel like when AI hears the word mockup, it's: oh, you want something generic, you want something blank, something plain as hell, so you can put whatever you want on it. And so you can spice it up yourself. And it's: yeah, yeah to a point. But designers like to kind of make something look a little.

How can you make a mockup look a little bit more sophisticated? So sometimes using the word mockup can kind of lock you and lock the AI model into giving you something generic. But maybe you can try something like using words like a high-quality editorial studio shot of a t-shirt, of a tote bag, of a cardboard box, of a letterhead, all that sort of stuff. And then all of a sudden you get something a little bit more bespoke. So maybe the box is now angled, maybe the background is a little bit more gradiated, all that sort of stuff.

So just one of those things that I learned the hard way. Now I'm spreading the knowledge in that sense. So a good thing to kind of keep in mind. And then another thing: distortion when generating a design onto a mockup can happen without the right prep. So making sure your design is the right aspect ratio for your mockup, or vice versa, helps save you headaches on things getting stretched or distorted.

So on the left here you can tell by the typography. It's kind of subtle but I might have stared at it for too long now at this point. But you notice that the headline is just a little stretched horizontally because the AI model is really trying to fill that space with the design. So that's kind of like not ideal. But on the example on the right is a little bit better in regards to how it's scaled proportionally across the space provided in the mockup.

So I'll be kind of teaching you guys how to kind of go about doing that and like avoiding that stuff so you don't have to kind of deal with the headaches and your designs appear on mockups as accurately as possible and then finally you can turn anything into a mockup. So if you have a crappy stock image but the bones are there, changing the photographic quality can go as far as getting you something that you can actually use. So a good example here is on the left. This is just a stock image of a woman in like a lab coat and then I went ahead and applied a styling prompt that added like a nice dramatic lighting and then I used another image node to just kind of focus in on the coat and all of a sudden you have a mockup for a biotech company or something with like a pin on it or whatever, and it looks nice and high quality and it's attention to detail like that really makes a difference in things like case studies, portfolios even. I know that's really really important.

Mockups and portfolios make such a big difference, Stuff like that. So there is so much that like an AI model can do. But kind of thinking about being given like a crappy stock image is actually almost like goldmine. Today with AI You can really really turn it into something amazing. If, how to, kind of if you have a vision for it, you can definitely get it there.

Okay, so now the quick part. Why nodes? And bear with me here, it kind of all ties in together because, yes, of course you can use your agency just like one shot, like 100 different mockups, but you give away a lot of creative control in that regard. So nodes kind of help bring that control back inand it brings a little bit more fun into the process as well. But I have a few pictures here you might be familiar with a few.

On the left you have like a whiteboard with post-it notes. In the middle you have like a conspiracy theory board with pins and strings, and then on the right you have an image of just like a mood board with images pinned up on the wall. So these are all like concepts that we're familiar with, like sticking things up and like making connections between them. So nodes, or basically nodes, are just kind of small. You'll see them on the canvas, but in context to the canvas, nodes hold like your AI models.

They're kind of like little windows that carry your images or your text or your video and you'll kind of see. You kind of see that in context. But bear with me here for a second. It's a natural progression of creative process and design thinking. So think of a mood board with pictures or a wall with post-it notes or even a conspiracy theory board with strings and pins.

We're more used to the concept of nodes and connections than you think. They also allow for repeatable workflows and visual systems. So making great visuals is one thing, but to systemize that approach and make it easy to repeat or pivot for a project or campaign is priceless. And then chat interfaces can only get you so far. Some of our first interactions with AI was through a chat interface, but as creatives have quickly outgrown that box and craved a more freeform way of making beautiful work without scrolling through endless iterations and finding the right version while your art director stands behind you or something.

So there was a question about what are nodes, and I think this might kind of answer that, but I'll take like 30 seconds in the chat. If you could type in your answer to this and if you've done this a million times, give me something different. Then What is a blue square plus a green circle? Equal? It could be a blue circle, okay, a ball in a box, that's good.

It could be maybe a green square, It could be anything in that sense. Basically, you kind of a beautiful backsplash tile- love that. A portrait: okay, Yellow triangle. I've got a lot of. I get a lot of triangle answers to this one, but that's kind of that's kind of something that I like to kind of bring up.

In regards to how nodes connect, You kind of determine how two things interact, or three things interact, or a bunch of things interact, all that sort of stuff. But let me put this in context a little bit, because this is not about circles and squares, it's about mock-ups. So how about this? What does the white bottle plus the design that you see here equal? It might equal: the flower designs on the poster is applied to the water bottle, or maybe the water bottle has the same headline as a design, or maybe the logo in the design is applied to the water bottle.

It all depends on what your intention is for that interaction between those two things. It could even be maybe you want the design on a t-shirt or maybe you want a specific flower on a t-shirt, but in context of how this might look in like a node kind of environment, you might have a design on the top and a mock-up on the bottom, and then you have these two lines that just drag out into one image node, and that is where you determine how they interact. So maybe you want them to interact by having the flower apply to the shirt, maybe you want the shirt color to change to the flower color, or maybe you want the flower from the first image to be embroidered on the sleeve of the second image. So the main thing that I'm trying to communicate here is nodes allow you to kind of determine how these things interact when you connect them together. So the nodes are basically what holds the thing you want to edit or control, and then the lines determine where they connect and how they connect, and then you kind of determine through prompting what that interaction is.

So then from here, what's nice about the node-based canvas is that you can kind of see the process from left to right, start to finish. So on the left you just saw like that little workflow that I just showed. But let's say you want to bring in a logo, so broaden the floor logo and I just place that right on top in the same vicinity, and now you can start to see how things connect from there. So let's say I want to take that embroidery shirt and now I want to apply the logo to the top of the shirt in the first example there on the right. Or maybe I want to change the shirt to black with the logo in white, or maybe I want to replace the embroidered flower with the logo on the bottom.

So again, whatever your heart desire, like you can communicate that to the AI via prompting, because if you don't determine the interaction between the things that are connected to each other, the AI model is going to kind of do it for you, which is not ideal. That's how you start to get like why I wanted the logo on the shirt, why is it on the sleeve, or why is the shirt white when I actually want the shirt to be black It's. You got to kind of identify what the things you want to change are and then voice that and determine that through a prompt. Okay, now let me show you what this looks like in the canvas. So before anything, I'm about to, I cut this short because I know like we want to spend as much time in the canvas together so you guys can kind of experiment as much as you can.

But bear with me, let me do a few housekeeping things here. So during our canvas time, first I'm going to show you how nodes on the canvas work, such as model selection, where to find these nodes, how to go like navigate the board, all that sort of stuff. And then I'm going to show you how to very easily take an existing design and apply it to a mockup. And then after that I want you all to try it out yourselves with provided mockups and designs for about five to 10 minutes while I do it with you and answer some questions, and then from there I'll show you how to change the photographic quality of mockups, as well as creating your own mockups from scratch and from references and then live experimenting. So there are things I didn't get to kind of do for the presentation, so I'm going to go ahead and do them live and experiment with you guys.

I'll show you demonstrations with the 3D node, I'll show you demonstrations with the image editor, maybe even adding motion to a few of these mockups, because motion mockups are pretty cool, all that sort of stuff. So that'll kind of be happening towards the end, as we have some more time left. So again, things to try out: apply design to a mockup provided or take a boring mockup and upgrade it with a technique provided. Try upgrading a boring mockup with a mockup that you like using text nodes, and try using a layer editor to design or to place designs in a mockup. On the left-hand sidebar of your canvas you're going to see this little toolbar and up here there's a plus sign and it says add notes.

So if I go ahead and click that, this is kind of just going to show you all the different options of nodes that you can kind of put onto your canvas. But, with that being said, we're going to focus mainly on text image, maybe video. I'm going to do text and image first. So I'm going to pop up an image node first, and this is what a node in FLORA is. So it's a little window, basically, that holds a lot of information.

So since this is an image node, it can use AI image models to generate images that appear in this little window. So let me just kind of show you the anatomy of this node really quickly. At the top of this node you'll see a few things. That's another little toolbar, and up here you'll have an option to kind of select what image model you want to use for this particular image node. I will say a few image models that you can kind of focus on today include Ideogram, Precise Edit 4.5.

That's going to be really great for editing. GPT Image is great for vector-based or text-heavy generations. Flux 3 is good too, and then Nano Banana Pro is great as well, and you're going to see me use a little of this Seedream 5 Pro for all that sort of stuff. So Nano Banana Pro is a pretty good all-rounder. We're all kind of familiar with it.

So you can either search it up in here, so Nano Banana, and it'll pop up there, Or you can scroll down and kind of see all the different AI model providers and then experiment with some models there. Next to the image model provider, you can also select the size of the node, so you can change the aspect ratio of your node accordingly. This is dependent on the image model that you choose, so just keep that in mind. Some models have more aspect ratios than others. Tools are basically a few little things that you can kind of run, such as color grading an image or maybe overlaying or drawing on an image, and you can also download image generation as well.

So that's kind of the anatomy of the image node. And then if I press T on my keyboard, that'll pop up a text node. Same sense: If I press I on my keyboard, it'll pop up an image node as well. So now that you kind of know the general anatomy of these nodes, let me kind of show you an example and show you in a context. So let's say, here I have an image of a T-shirt, a guy with a T-shirt, and I'm actually going to do this live to kind of demonstrate this fresh.

I'm going to move this over. So let's say I have a guy in a white T-shirt- the same example I showed you guys before and then a flower image. When I hover over to the right side of an image there's this little plus sign. I can go ahead and click and drag that out. If I let go, then I can select another node that I want that to be connected to.

So I went ahead and selected an image nodeand now I'm going to go ahead and connect the flower image as well. And you'll know that they're connected because they show up here in this little thumbnail areaand they show up in the order that you connected them into the node. So you can either flip them around if you need to, depending on how you kind of prompt it. But once you have these two reference images kind of connected into this image node, you can start to kind of prompt, straighten the image node. So, for example, here, as easily as I can speak it, I can prompt itso I can say something: apply the flower in image two to the shirt in image one.

I'm going to use- I'll use Nano Banana Pro for this. So when I have my image node and I have my prompt down here, you'll see a little option that says run. This is going to generate that prompt for you. But you'll see here there's also an option to generate a specific amount of variations from that prompt. So this is pretty straightforward.

There's not much room for interpretation. But sometimes you want to see options, maybe if this was a little bit more of a complicated prompt, such as, apply the flower in a abstract way onto the shirt- And I'm actually going to say that, apply the flower in the second image to the shirt in the first image in an abstract way. Just gonna correct that. That kind of has like a little bit of room for interpretation. So maybe I want to see four variations of that.

So when I hover over this white arrow on the bottom, I can go ahead and click this. My magic number is four. You can technically go up to 20 variations, but I tend to just do four. And if I don't get what I want out of four generations, then I'll try to either change my prompt or change my image model. So I'm going to go ahead and click, go, and when I do that, then four notes pop up onto the canvas.

So from here, as you can see, I have four pretty, pretty different interpretations of what it's like to apply flower abstractly onto a t shirt. They look fine, but we're not here, for fine Designers are never satisfied with its fine. We got to make something better right. So I'll show you that next in a little bit. So here I have an example of a wheat paste poster mockup on a brick wall and, as you can see, these are all white against a brick wall.

It's kind of hard to see the distinction between each of the posters. So for my piece of mind, what I did here was I actually attached this image into another image nodeand I said: color every other poster in green. That way I can easily visualize and refer to each section of the poster area by a color, because now I have two other poster designs that I want to start to apply into that mockup. So I'll go ahead and re demonstrate this with the same prompt here, but I'll move this out of the way. So let's say I have a mockup up here.

I'm going to drag that out Again. If I hover over the right side of an image node, click and drag and select image, I'm going to go ahead and make sure my aspect ratio is set to auto, and setting your aspect ratio to auto will automatically adopt the aspect ratio of the image that's attached. So that's kind of why I just want to set it to auto and I'm going to go ahead and attach my two other designs here. So now I'm going to prompt it with: replace the green posters in image one with image two and replace the white posters in image one with image three. So you might see me kind of do like say things like in image one and image two, image three.

Personally, that's kind of the easiest way for me to quickly reference, like how images that are connected into a node interact. So, as you can see here what I mentioned earlier, the nodes that are connected into another node are ordered in regards to how, basically how you plug them in. So since I plugged in my mockup into this image node first, that's image one- And then I went ahead and plugged in my other designs here, If you wanted to swap any orders you can do that as well, but I'm not being too precious here so I'm pretty okay with that. So I'm just going to go ahead and hit generate. The more you edit an image in AI, the more it degrades in quality.

So the finer details about an image start to get melted together and smoothed out. That is so funny. Replace the green posters in image one with image two. I'm going to give that another try. I'm actually going to use Seedream 5.

But what's nice about this model- Ideogram 4.5, Precise Edit- is that it does a really good job of maintaining quality through generative processes. So if I actually select these two images and if you ever want to compare two generations together, you can select them both. And when you do that at the top toolbar there's a little option called compare images. If I click that you'll see them side by side. But I'm also going to go ahead and go to the top left corner and go to slider, and what I can do with the slider is actually see how accurately the model carried over details.

So the coat texture is almost perfect. Look at that seam in the coat on the left and then after the generation it's still there. If Nano Banana did this, that would be gone. JK, maybe it wouldn't. AI is kind of getting better every day.

But really, really love this model for precise editing without losing a lot of fidelity. Let's say you have a vertical poster but you have a horizontal mock-up, so you can take a design, plug it into another image node. You can set that image node to the aspect ratio that you want. So in this case it's really wide. So I set it to like 21 to 9, but I'll go ahead and try this at 19 by 16.

And for this type of thing I would only recommend a GPT Image model, specifically 2.5 Sunburst. It's their newest model because that provider does really well with layouts, with vectors, with text, all that sort of stuff. It may not be perfect- I mean ideally, in a perfect world, you would redesign this for a horizontal setting so you have more control over how the type is set, You have more confidence about the type, fidelity, all that sort of stuff- But in a pinch you can do something like this. So once you get a kind of generation like this, you can then plug that into an image node with your mock-up and then say, apply the design in the second image to the blank wall in the first image, and what? It's OK, but it's not quite long enough.

And plus what's going on with the logo up here? That's not right, That's not going to cut it. So this is a little like I kind of like I took a creative stab at this, but what I did from here was I basically plugged this into an image editor, and if I wanted to do that. I can just go ahead and drag this out and on the bottom you'll see an option for image editor and if I click this little icon in here it'll put me into kind of like an basically a mini Photoshop in FLORA. But if I select the layer on the left, there's a little draw tool on the right where it says edit layer.

I'm going to go ahead and select the draw tool and I'm just going to clearly- oh, not in black, I'm going to do this in red- I'm going to clearly draw a line at the edge of this and then save, and what that's going to do is basically act as a reference for me to then plug that image editor into another image node and say: resize the back poster proportionally so that the edge highlighted in red meets the end of the composition. This will make the overall back poster wall taller in proportion, keep the design and everything consistent, and for this I use Ideogram, Precise Edit again, and the results were pretty good, so happy with that. If you find yourself having to kind of bend backwards for something like this, there are options. If you need to draw a line, use the layer editor tool to kind of get there and then kind of speak your way through, adjusting that size according to the mock-up. I have that same mock-up, but instead of two posters I have four.

Now, instead of prompting my way there to kind of apply these mock-ups or apply these designs in the mock-up, I actually have this image plugged into an image editor. So if I go ahead and re-demonstrate this, I'm going to plug this in image editor. I'm going to go in here. This will work for some mock-ups better than others. It depends on how complex the mock-up is.

Since this is pretty simple, the layers are kind of like self-explanatory. This is okay. But if you have an image that's like very complex, with a lot of people, a lot of products, a lot like a busy setting, this might not be as powerful. But when you go into image editor with an image plugged in on the left-hand side in the layers window, there is an option to split image into layers. I'm going to go ahead and select Speed Dream, or, sorry, Speed Dream, Seedream 5 Pro Layer Eyes, and what this will do is generatively cut out each part of the image and place it on its own layer.

So I'm going to let that generate, but show you what it looks like here. So, as you can see, the shadow is on its own layer, The woman is on its own layer, All the posters that I have plugged in are on its own layer. I even have the brick on its own layer. All these sort of things start to get split up, and so now you have a little bit more control over where things go, how big they are, all that sort of stuff. So if I kind of just hide the posters, this is what the composition looks like after doing the C Dream layerizer.

So, of course, some of these are not, as I don't know what's going on here. But it's fine, because what's going to happen is I'm just going to place my designs right over it. So these designs can be. If I wanted to, I can make them smaller. If I wanted to, I can also just kind of fill a little bit more of the brick wall, all that sort of stuff.

So I have a little bit more confidence about, yeah, the text isn't getting distorted. I have a little bit more control over how things are looking, all that sort of stuff. And then I can delete this last one up here If I don't like the shadow, I could remove it If I think it's too dark, since the shadow is on its own layer, I can go ahead and just lower that opacity, if anything, or maybe I want it full, I don't know- depending on how dramatic you want to be, and you can also set the blend mode of a particular layer. So maybe I want the shadow to be multiplied, Maybe I want it normal. What does it look like if I put this screen?

Yeah, that's not really helpful, so I'm just going to put it to multiply. But yeah, what's nice here too is actually I didn't try this- but you can also generatively edit a specific layer in your layer editor. So I have this woman in her trench coat selected. I'm actually going to go ahead and say, change the coat of the woman to hot pink, and so now that is the only layer that's going to be generated with that edit. So if you want to kind of try something like that, you can do that as well.

What you can then do is attach that image editor into an image nodeand then you can prompt something as simple as blend the posters seamlessly into the brick wall as if they are wheat pasted onto the surface, and then your rough composition all of a turn, all of a sudden turns into a pretty realistically mocked up composition. I have an example where I have basically an unedited version of the mockup I was just showing you and I have that plugged into a color grade image action. So how I got there is basically when you hover over an image at the toolbar there's an option for tools and if you go ahead and click that down here there is a- there you go, a color grade image option. So if you press that, this little purple node will appear and think of actions as creative tool nodes. So while you have image editor, which is like a much more robust kind of like interface where you can like draw, you can add text, you can move things around, actions are like many processes condensed into one specific node.

So this is just an action node dedicated to color grading an image. So if I go ahead and double click into this action- let me hide the chat here There- you go Over here on the right, you're going to see a few parameters that you can start to mess around with. So you can do adjust warmth, contrast, saturation, all that sort of stuff, and these are non generative processes. So there is no AI model that's adjusting the warmth here. There is no AI model adjusting the contrast.

This is literally as if you opened it up in a color grading tool such as Photoshop, just kind of easily done on the canvas. So you don't have to kind of leave if you don't have to. So really, really easy little thing that you can do in FLORA if you just need to tweak a little bit from that image. And if I kind of compare these two, now and again, if you want to compare two images that you just generated or that you edited, you can select them both, go up to compare images up here and you'll see them side by side. So I think the one on the right looks much warmer, It looks a little bit more like daylight, while the one on the left looks a little drabber.

So I think we prefer the one on the right. On the left I have just that blank- we call it that blank mockup- And I have that attached into an image nodeand I simply again just kind of prompted it with apply a harsh direct flash photography effect to the image attached and I'm pretty happy with that. I think the show is a little intensebut again, with the layer editor I am kind of adjusted that from there. But here that was kind of just as simple as speaking it. High direct flash effect- Great.

So if the photo effects that you want to apply to the image is something you can easily say- not much room for reinterpretation, not too many details- you can just go ahead and straight up, prompted in the image node and you'll get something pretty close. Another kind of version of this is: I went ahead and apply, applied that it mockup into another image nodeand I said something like relight the image to make it seem as if it is a sunny day outside with slight tree leaf shadows casted on the wall. I'm yeah, I'm yeah. Okay, that's kind of cute. It's looking very autumn.

Maybe I might want to prompt in some autumn leaves on the sidewalk, I don't know. But as easily as I can kind of speak it, it kind of reinterpreted what that might look like here. Now, this one is pretty interesting. If this isn't your first time kind of going to a workshop that I lead, you might be absolutely tired of me saying articulation and prompt creation. But the day I started at FLORA, this has always been like my ride or die workflow.

In regards to using a reference image like this is if I had to pick one workflow for the rest of my life, it would be this. So let's say you have a nice image on the left, you like particular parts about it in regards to the photographic effect of itand you want to kind of like bottle that up into a prompt and have it apply to other images. So let's say you have an image such as this. I have this plugged into a text node, and this first text node is where I say something. I like the photographic effect of this image, specifically the shadow contrast and glow of the light.

Can you analyze these qualities? Ignore what is being depicted. There should be no mention of details specific to the composition. I only care about the photographic quality of the image. So there are a ton of different ways that you can kind of like garner information from a reference image.

In this case, I see this image and I see a photographic quality. Like you might see this image and you might say I like I see this image and I see a good white t shirt mock up. I see this image and I see a good candid photograph opportunityand you may want to articulate those details about the photo. In this case, I'm only focusing on the photographic quality. So use this first text node to kind of analyze those details.

There is no prompt creating here, I'm just asking for an analysis. And then the second text node is where I say: condense these details into a prompt I can use to apply the same photographic effect to another image. And so then it goes ahead and writes me this comprehensive prompt. It's apply a 30- 35 millimeter color negative effect like film, look to this image, keep the subject in composition and transform the lighting and tone, blah, blah, blah, blah, blah. Amazing, it sounds great.

So then what I do is I take that prompt into another image node with the mock up that I want to apply that effect to, and then I get that same sort of effect applied to my mock up. Now, this is pretty extreme, this is really. I don't know if I would actually kind of use this for a case study, but it's a really good example of taking a photographic quality from one image and applying just that essence, just that effect, to your mock up in a very nuanced way. So if you have a very dramatic type of- like gritty sort of branding system, a mock up like this might actually work. But yeah, I just kind of wanted to find an example that would show you something like this to this effect.

So this I'll show you how I actually got the prompt, the mock up that I'm was kind of demonstrating for you earlier in the workshop. So I opened up a text node- I'll do this again- and then I paste it in Again. I type this all out. This is me literally typing into the text node and prompting it. I have a two to three poster design or a two by three aspect ratio poster design that I need to mock up as a series of wheat paste posters.

The mock up needs to be straight on a brick wall in an urban street and look like a high quality, premium design mock up. The wall shouldn't look too messy. I need blank white space for my poster designs and I like motion blur people in mock up. So I just go ahead, went ahead and added that inand then I asked it to write me, write me a prompt to do this. So once I have that general sense of what I want, I kind of just go ahead and generate and I'll get a really good, comprehensive prompt here.

The common thread behind my workshops is: use text notes to write the prompts for you. So went ahead and did this, connected that into an image nodeand this was one of the generations that I got. I'll go ahead and generate a few more just to kind of see how they come out. But basically once I generate that I was okay, sure, I liked it. But I noticed that there were gaps between each poster.

So I went ahead and connected that into another image nodeand I said: put the posters right next, right next to each other, so there are no gaps, and fix the peeling corners. So once you kind of generate something off of a prompt, if it's still not perfect, just extend that baby into another image node and then edit from there. So that's really easy way to just kind of get a good skeleton and then edit, make a few adjustments to kind of make it a little bit closer to what you want Up here. The same way that I showed you the prompt creation, articulation and prompt creation, I'm kind of doing the same thing here with this photo effect. So before I showed you what I did for that t-shirt image where I liked like the vintagey effect of it, Now I see this image and I'm oh, I like a lot of things about this photo, so I grabbed this photo, plug it into a text node and I say: articulate the photographic art direction of this image.

I specifically care about the shadows and highlights and the contrast to the background. Got that articulation. Use a text node to condense the details into a system prompt I can use to apply the same photographic style to another image. And once I did that, that's how I got this upgraded science image here Really effective. And then, since it's a little zoomed out, I kind of wanted to focus in on the lab coat.

I just kind of plugged that image into another image nodeand I said: focus in on the lapel of the lab coat for this image. I didn't ask it to add a pin, but like when I saw this, I was actually not bad, I'm pretty happy, but I could probably just make a 3D model of another logo and pop that baby in there. Let's go back to this guy here. So remember when I kind of used that articulation and prompt creation to garner the photographic effect of the image. Now I'm taking a few steps back here.

I like this image. It could be a good basis of a good t-shirt mock-up, but I don't want to use this photo because for the sake of presentation. I have images that I'm okay sharing with you guys and sharing to the public. But in your case you might find something on Pinterest that you might, but you don't want to copy it because, again, we're not trying to copy an existing mock-up, we're trying to make something new. So in this case I see this image and I'm that could be a good shirt mock-up, but I don't like that exact composition.

So let's see what we can do here. So I have this image plugged into a text nodeand here I'm prompting something. Write me three prompts that would create a blank shirt mock-up, loosely inspired by the image attached. Number each one, two, three, no dashes. The pose can change, the model can change and the camera angle can change.

This is where I can. This is where I outline to the to the text node: Hey, these things can change, These things cannot change. Keep these consistent. Whatever it may be, I'm being pretty loose here. I want to see more creative outputs because I'm I just I'm exploring here.

I'm not being too precious about it. So once I have those three prompts again, I kind of attach it into a batch node so I get each prompt individualized and then again these batch notes kind of just like disappeared. And then I start to see generations based off of those prompts that are pretty close to the original, but not exactly so, as you can see. Another thing to kind of keep in mind is this image is not plugged in to this right image node, so it actually has no interpretation of that original reference image at all. It only has the prompts that were created based off of it, and that's it, because a node only has information and context of the things that are plugged into it, not things that naturally live in the canvas.

So just something to keep in mind. This is me intentionally not plugging a reference image into a generation, because, again, I'm not, I'm not trying to stay too close to the reference. I already kind of got what I needed there and I was about to try this out, actually before we started here, but I had a. I created this during one of my CPG workshops or webinars a while ago and I did the same thing. I plugged that into an image node- sorry, a text node- And I said: write me 10 prompts that would create a blank mockup loosely inspired by the image attached.

Number each one through 10 node dashes. The composition, positioning scale can change and the camera angle can change. So this text node spat me out 10 prompts. I'm going to go ahead and plug that into a batch nodeand I'm going to go ahead and plug this into an image node. I'm going to use cdream for this, I'm going to put my aspect ratio to like four to three and then I'm going to hit run.

But again, these are some examples of that first batch: Pretty like. I got the vibe. The vibe is there that I liked from this image. I'm pretty happy, like I can pretty confidently be. Yeah, that's a.

That's a pretty decent like shirt mockup there if it was for like a food and beverage or CPG product in that sense. And I got there with just like a loose reference and a few prompts: Nice, Okay, some of these are actually pretty nice. Oh, wow, Okay. So I have a mood board here. I have a screenshot of a Pinterest board here.

I did I didn't want to just like bring, bring in one specific mockup here, I just wanted a general. How do I put into words the aesthetic, like the vibe of this Pinterest board that I like? So I just brought in a screenshot of all of it. You can bring in one image. You can bring in 10 images.

You can bring in one screenshot, screenshot of 50 images. But what I did here was I took that screenshot, put it into a text nodeand again I just kind of asked it or I said: attached as a mood board of mockups I like Generally, I like the style of compositions, the way the mockups are framed- simple, minimal, etc. Can you articulate these qualities and directions? Went ahead and did that and then I use the text node again to condense those details into a prompt I can use to style other images in the same sense. I'm asking it specifically to start off the prompt that it's going to write for me with quote: restyle the attached image accordingly to act as a high quality designed mockup.

There you go. So once I have that prompt written for me, I took one of, like my water bottle mockups and I just kind of applied that there and got something like this- Not the best, I've probably regenerate that there, but what I did was I went ahead and took this. I took this workflow and took this prompt and created a technique out of it for you guys to kind of mess around with. So I think I mentioned I touched on techniques earlier, but techniques are just condensed workflows into green nodes that look like this. So this is a condensed version of that mock, that workflow I just showed you.

And if I hover over this technique and I go to the top where it says view workflow, you'll see that within is just like one text node with that prompt that was written for meand I made the technique so that you would get four outputs on the other side based off of one input. So how I went about doing that- I won't get too nitty gritty about itbecause again we're just here to kind of like learn how to make mockups and whatnot- I went ahead and isolated this prompt here. So I don't need that mockup image. I mean, I don't need that mood board image anymore, I don't need the articulation anymore. The real like gold of this workflow is this prompt that was written for me here.

So I'm going to go ahead and copy that into a text node as such. And I went ahead and locked this text node because I don't want this text node to regenerate every time someone runs the technique. I just kind of wanted to say static, because when I plug that text node into an image node it automatically populates there. So what I did was I just kind of did this four times because, again, when I'm creating a technique or when I'm creating a workflow for other people to use, I'm very strictly understanding what I want them to plug in and what I want them to get out on the other side. In this case, I want them to plug in whatever image they want of a mockup of an item whatever, and on the other end, I want them to get an upgraded version of that mockup and I want them to get four variations of it.

So that's why you see four things here. So once I have that, I can go ahead and turn this into a technique. So what that would look like is up at the upper right corner there's a build technique option and the first thing it asks you is: what are inputs? Go ahead and click that- my input on the left- And then it's going to say: what are your outputs? Now, go ahead and do that, because I want all four of these to be my outputs- And then this is what the technique will look like.

You can name your techniques. So in this case it's mockup upgrader. You can add a description. I already did this, so I'm not going to do that there You can publish a technique to the workspace or privatelyand then you would hit publish. But here's an example where I took one of my product, like one of my CPG projects, plugged it into that technique and then I got some good mockup upgrades.

I was pretty happy with some of these. Honestly, This one's really nice. All right, Thank you guys. I just put in the chat a link to that event tomorrow night. Feel free to come on through.

We would love to have you guys all there. It will be open bar, so it'll be a great time and lots of creative, awesome people there. We will offer these labs the next four weeks in partnership with AIGA New York, so feel free to come on next Tuesday from 6 to 8 p M Eastern time for the next lab, as Kat mentioned, and until then we have lots of webinars and other FLORA Labs up on our FLORA's Luma. So feel free to tune in there and we will see you guys next week.

Thanks for being here. Thank you guys. All right, take care Thanks.