Episode 4: Innovation in Action with Fabric Genie

Want ideas on how marketing professionals are actually using AI in their daily life? Listen in to digital markerting expert, Will Francis and SEO consultant Danny Richman chat about what's new in how AI is affecting marketing. 

In this episode, Danny talks about how he built a pretty amazing AI-powered pattern creator for a well-established British fabric company. The first of its kind, this is a pretty unique example of the latest technology paired with a traditional industry, and it's working out well. You can try Fabric Genie our for yourself on the website. 

You can also watch our earlier webinar with Will, Danny and Clark Boyd where they dig deep into what ChatGPT really means for marketers.

“One of the kinds of things I keep saying about AI is, rather than just trying to find things that it can do, that humans are already doing, and just trying to replace humans is let's find things we can do with this technology that are not already being done. ” Danny Richman

Use AI to supercharge your digital marketing activities

If you're looking to up your digital marketing game through understanding the full benefits of AI and using the best tools, sign up for our short interactive AI course today. 

Video transcript

Will Francis  00:00

You've just helped launch something called Fabric Genie. What's that? Tell me about it?

Danny Richman  00:06

Yeah, well, bit more than just helped launch it, I built the thing. Well, I built all the back end of it, and a very talented front end designer, called Seb Kay built, all the lovely front end to it made it look beautiful, and all this kind of thing. So yeah, this was a project that I did for a client that I was already working with. So I've been helping them with their marketing for quite some time now. And we were just trying to look at things that we could do, you know. One of the kinds of things I keep saying about AI is, rather than just trying to find things that it can do, that humans are already doing, and just trying to replace humans is let's find things we can do with this technology that are not already being done. And so I became really interested in all these image generators that were coming out Stable Diffusion, Dall-E, Mid journey all coming, you know, doing all these, creating all these images. This particular company, this is a soft furnishing retailer, called the Mill Shop online, they sell curtain fabric, upholstery, fabric, line, fabric, that kind of thing, been going a long time very successful business.

Danny Richman  01:23

And we thought it'd be really interesting to provide a service on their website where visitors could go on there and just describe the fabric that they want, or upload an image of their room. And then it would create fabric designs based on what you know, whatever would best suit, the room, the picture of the room that they've uploaded. And so yeah, we came up with this idea, I think it was around about 12 weeks ago. And so we went from idea to launch within the space of 12 weeks. And yeah, it's gone down a storm, people are now using this. It's only just been launched. But we've had now 1000s of people going on, they're creating designs with this tool, then they order, it's one click from getting the design to then ordering a sample of it on fabric, so you can touch it and smell it and hold it in your hands. And then one more click to then get it made up into beautiful curtains, or blinds or cushions or loose fabric or whatever you want to do with it. And as far as we're aware, no one's done this before. And aside from the fact that this is a really interesting product to offer people that you can now have something that is completely personalized to your tastes, your requirements, your room, you can literally have anything you want on this fabric. There is also a marketing play to this as well, in that it has already created quite a lot of coverage around the media. So, the soft furnishing trade press, the AI media sector publishers, and also people that write and talk about ecommerce. And so, obviously, that all increases brand awareness for my client, the Mill Shop, and also drives links into their website, which then has a positive impact on their SEO. So there are various strands to this, all that are beneficial to their business. For a project that really didn't take that long and cost very little to put together.

Will Francis  03:47

That's interesting. So it basically it produces tiled patterns, right? I've used it and you when you get this the pattern ideas that it's hard, they're all they would all tile so they're all repeating patterns. So I can see there's obviously some, there's some instruction that's baked into it. Obviously one of them is to tile the patterns. Where else is it getting its source information from? Because I one of the last things I did with it was I made some William Morris style patterns with very specific plants and flowers in and it totally nailed it like it, you know, it? It was brilliant. So how does it know what that looks like?

Danny Richman  04:26

Now we have, it doesn't just literally take whatever you type, and then turn that into a print design, as you rightly guessed: there's a few little things that we've needed to kind of bake into that to get the best results we can out of those designs. There's also had to be an awful lot of moderation put into that process, because obviously making something like this available free to the general public, you're gonna get an awful lot of people that are going to use it for weird purposes we've already had, trust me, we've had weird requests coming in. And so yeah, so we have to track all of that as well. And one of the other challenges we had is, you know, it's one thing using an AI Image tool and typing in a description, and then you wait, and then you get back an image. We've had to make this scalable, so that if we get, you know, 2000 people or requesting an image all at the same time, they don't want to be waiting for hours for that image to be generated. And so that was another kind of technical challenge with dealing with this. But what's interesting is that actually, the whole back end process of this thing was built on Zapier.

Danny Richman  04:26

Well, it's really no different to using a tool like ChatGPT or, you know, if you go and use like Stable Diffusion or one of these tools, you know. It's a language model is essentially what it is. So they start off by training the model on huge amounts of data, in this case, images. Millions and millions and millions of images will have been fed into this model, along with a whole load of text so that it understands the concepts of what you're typing in. So when you type in something like in the style of William Morris, what does that mean? It has to have an understanding of that concept of William Morris, you know, or flowers of roses or whatever it is you put in there. And then just like ChatGPT, it's trying to come up with the most probabilistic response to the request that you've put in.

Will Francis  06:39

Wow. I was going to ask that, actually, is this all built on publicly available, kind of fairly inexpensive tech that's stitched together just in a very specific way?

Danny Richman  06:50

I mean, there's very little code involved. I mean, obviously, there's a limit to how much I can talk about behind this, because this is aspects of this that are commercially sensitive. But there's very little code actually involved in this whole back end process and what code there was - there was some code needed - but all of the code that we needed was generated by GPT-4.

Danny Richman  07:17

We just instructed GPT to give us the code that we needed, put that into a code step in Zapier. And yeah, this whole thing is built on something like about I think something like about 80, different zaps that are running continuously throughout the day and night.

Will Francis  07:34

Okay, so you're definitely on a premium plan for Zapier?

Danny Richman  07:40

Yes. Team plan, I think they call it Because obviously, it burns through quite a lot of tasks.

Will Francis  07:45

Yeah. Wow, that's so interesting. I mean, I think in the product innovation community, I can see why and in the AI community, just the idea that you can stitch together readily available things into something original is captivating in itself. And I think that's the opportunity for any business. I mean, your client aren't the obvious AI innovation business.

Danny Richman  08:14

That whole industry, believe me, that whole industry has just been so resistant to any kind of innovation. You know, I had some involvement with the soft furnishing industry about 35-40 years ago. And this is the first time I've had any kind of interaction with any company within that industry. Nothing's changed. I mean, really, nothing has changed. And so, that was another element of this. It was just kind of, well, I've been away from that industry for a very long time. My journey has taken me down a technology route. So it's quite nice to sort of bring my experience back into that industry and see what I could do with it.

Will Francis  08:59

Will you be able to share any results from this at any point?

Danny Richman  09:04

Yeah, I mean, so. So it's very early days yet, you know, we certainly have found that there's a lot of people using it. We've had some really positive feedback from people. I think, generally, people have been very surprised by the quality of the designs that have been generated. I think a lot of people's expectation was that the designs would just be a little bit kind of cheesy and naff. And, and that's not been the case at all the quality of the designs - and that was really important for us - was to make sure that we were only doing this with a technology that was capable of producing really high-quality output that you would actually want to buy, that you would actually want to have in your home.

Will Francis  09:47

That's a very high bar, isn't it? Something you'd actually want to print on a sofa or a pair of curtains. Because that's what I was looking for in those patterns was kind of weird, you know, nightmarish sort of Uncanny Valley anomalies. Like roses that were didn't really look like roses, and were weirdly kind of distorted, but didn't really get any of that. So it's very interesting.

Danny Richman  10:08

You don't get that at all and the other thing, the other element of this as well is that it opens my client up to a whole new market, which is in the world of interior design and commercial interior design, one of the things you can do with this tool is you can be very specific. And so you can even put in exact Pantone color numbers and color references. And say, you know, I want it in Pantone 1046 Brick Red. So if you're working on a commercial project, where you're working with a very specific mood board, you know, and kind of theme that it's got to comply with, you could spend weeks trying to source fabrics for a project like that, you know.

Will Francis  10:54

That was one of the things I wanted to ask you about really. With these tools, it's so hard to know how much freedom to give the user because on the one hand, you don't want to give them too much thinking to do: if it's just an empty text box, like right, your prompt here for your fabric. Most people are like, I know, I need buttons and levers and sliders and stuff. But then of course, if you really oversimplify it, you take the power and the interesting-ness out of the results, potentially, I mean, thinking about that project, but also in general, when building these kinds of tools, what do you think about that? How do we find that balance?

Danny Richman  11:33

So I don't know if you notice when you went when you tried this out for yourself. There's kind of different modes, there's different ways of using it all. So we've got, there's a kind of a theme mode, which is where you've got multiple fields. And it's basically just a textbox. But what we're doing is just kind of prompting you to make it easier for you to understand the different elements of the design. So you can specify the overall feel and look of it, you can specify the foreground colors, the background colors, the individual elements that you might want to have in that design. And it's really, it's just nothing more than a kind of a guide to help the user think about what they might want to see in the finished image. So you've got that option, then there's a kind of just a text box. So if you just want freeform text, and you're feeling, you know, I don't want to be constrained by all these different fields, I just want to be able to type out exactly as I want it, you've got that option too. And I think I mean, time will tell, but I think that's a good way of approaching it is for people that need that helping hand give an option for them to do that. And for the people that don't want to be constrained, give them the option of just being able to use it the way they want.

Will Francis  12:51

And you can upload a photo, which I thought was really good.

Danny Richman  12:55

So there's a couple of options there, you can either just upload a photo of your room and say I'd like a geometric print. And it will use the colors that you've uploaded on the room image. Or if you've got a sketch, or you've got an existing design, you can take a photo of that, upload that and it will then be influenced by that design. So, it's early days for this project, to really know whether we can consider it successful or not. But it certainly generated a lot of interest. It certainly generated a lot of media attention.

Will Francis  13:34

I think that's the thing, these projects, isn't it, it's by no means a kind of a cheap tacky newsjack-y type tactic, in my view, to do these things for the PR, I think that's a really valid reason to do these things to appear to be a first mover and an innovator. And to be trying things that are interesting and disruptive in some way. Right?

Danny Richman  14:03

It's also just a great thing rather than having to go to a website and start trawling through hundreds and hundreds and hundreds of different samples of fabric looking for the one that you want, you know, which is a time consuming and time-consuming process. If it can do it to that quality? Why not just create it? And of course, this has implications outside of the world of soft furnishings and fabrics, you know: that level of personalization and customization using AI can be applied to so many different products.

Will Francis  14:39

It can. I think that's the next big wave in in commerce is increased personalization, clearly, both in things like that way where it's print on demand: any product that's print on demand, like fabrics, you know, like notebooks and home furnishings and things Yeah, but also Where there's a limited number of items or SKUs within a product inventory, like a big fashion retailer is still, you know, matching that shirt with those trousers and those shoes. Everybody still wants a personal stylist. I don't know about you, I bring my kind of personal desires to these conversations all the time. I would love a personal stylist personally, I would genuinely love a personal stylist. And if that cake comes along in the form of AI I will spend more money on clothes. Fact.

Related Articles

• AI in Digital Marketing - The Ultimate Guide
• Webinar: What Does GPT Really Mean For Marketers?