Curious about how AI tools can supplement your copywriting efforts? Dave Feinleib, CEO of It’sRapid, and Oscar Kaszubski, Chief Growth Officer at FirstMovr, dive into content creation leveraging leading AI tools including Chat GPT 3.5, Chat GPT 4.0, and Dall-E.
Transcript
David Feinleib
Hi, I’m Dave Feinleib, founder and CEO of ItsRapid.ai, and I’m joined today by Oskar Kaszubski, Chief Growth Officer of FirstMovr. Welcome Oskar, and great to have you on the show.
Oskar Kaszubski
Hey, Dave. Well, thank you for the warm welcome and thanks for having me.
David Feinleib
Well, today we’re going to be talking about a timely and interesting and exciting topic, which is generative ai, and we put together some really interesting approaches and some prompts that we’re going to be using across a variety of different tools. Oskar, tell us a little about these different tools, if you would. And before we dive into the examples, what are we going to be looking at here?
Oskar Kaszubski
So, we’re going to be looking at generative AI tools, and we wanted to focus on two things: One of them is text generation as well as image generation. So we’re going to be looking at chat GPT 3.5. We are also going to look at what is the difference with version 4.0. And then for image generation, we’ll look for the best in class, what is available currently. So we’re going to look at mid journey version 5.1, as well as DALL-E 2. DALL-E is an image generation that is actually built by open ai. And then we are also going to be looking at a stable diffusion version SD xl, which is a stability AI tool.
David Feinleib
Love it. Amazing. So, we’ve got a variety of different options, and we’re going to start out here with Chat GPT, which is well known. And for this one, we put together a prompt around walmart.com with Doritos spicy sweet tortilla chips. So I’m going to fire this one off with Chat GPT 3.5. And let’s see what, we’ve got a pretty extensive description here. We’ve got our product copy experience, a taste sensation like no other. We’ve got our feature bullets, which are great. Talks about the bag size, the bold and spicy flavor. What do you like about this one, and what do you think could be improved?
Oskar Kaszubski
Well, number one is, what I like about it is that it’s written in a very, kind of a compelling way, right? Which is like, it creates, it’s a very creative copy that actually a lot of consumers can relate to where I, where I do think it needs a little bit of a training, and that’s why prompts like this are not enough. What are the limitations of the retailer? What are the specific search keywords that need to be included in that copy? I would also maybe add some of the call outs from category management. It’s like, hey, am I trying to increase household penetration? Am I trying to increase frequency? Am I trying to increase basket size? Right? So we could actually fine tune this copy quite a bit to get to the right output from the perspective, but it’s definitely a great start. But just think about it as the algorithm is not trained necessarily on the constraints of the retailer. So it’s very important that you actually add additional context to it in terms of specific character sizes for each of the,from, the title, the bullets, et cetera.
David Feinleib
Yeah, yeah, a hundred percent. Now, if we were to take this one and even try regenerating it, we’d get something a little different. How does that work? When, when it regenerates these, should we expect to be seeing something different or, what, what would we wanna do differently here if we were a brand thinking about this?
Oskar Kaszubski
I mean, what I think is, each prompt is, has a different seed, right? So within image generation, which we’re going to look at next, you can actually specify a seed so you can actually get to the same image. But, AI, text generator works in a, in a very similar way. But I think here really the key is to be the prompt engineer and, and start actually adjusting some of those keywords in the original prompt to be able to actually get to the right level of output.
0 Comments