Elevate Your Content with ChatGPT's Enhanced Image Generation: A Guide for Accountants

Let's get into our subject for today, which is AI image generation. So what I'm going to cover is we'll talk a little bit about what is this new tool, what can it do, how can we use it? I'll talk about some use cases, I'll talk about some tips for getting better images. And so we have got. We've got a lot of stuff to cover in this session. So by way of introduction, what I'm specifically talking about is the new image generation tool that OpenAI released. They released it into ChatGPT on March 25, 2025. So as I'm recording this, it's brand new, this tool. And what it's done is it's replaced Dall? E Dall? E 3, if you use that, and I know many people who love ChatGPT also use Dall E3. The problem with Dall? E3, in all the various tests and rank of what is the best and worst AI image tools, Dall E3 came right at the bottom of the list. It was okay, but much better with tools like Midjourney, for example, was one of the best ones. But the reason why I think Dall E3 was so popular, despite the fact it was pretty poor compared with other tools, I mean, it's still amazing when you think what it can do compared to what we were aware of two, three years ago is extraordinary. So the things that we can do these days. But it, it was pretty poor. The only reason I used it was because of convenience, because it's inside Chat GPT. I'm using ChatGPT every single day. And so if I want to do an image, it was much easier just to go to Dall E3 rather than to go and log into something else, like Midjourney. I have a Midjourney account, but I just found it. I'm lazy. I want all the stuff in the same place. Now, if you do like Dall E3, it's kind of not gone away, it's still there if you want to access Dall E3. Just so you know, if you do go into ChatGPT and if you go to Explore GPTs, then when you explore GPTs, you can search for Dalle 3 and the OpenAI version of Dalle 3 will pop up, and you can use that if you want to. But even that, it does say a new model is rolling out in ChatGPT, which it has. It did on the 25th of March. So that's the big change that took place on the 25th of March. Dall E3 is gone, but you can access it via the GPT. At the moment, they've got the new model. Now the new model you can access through GPT4O at the moment, that's bound to change because they keep changing stuff. And if I just show you how you access it, because one of the things I noticed is they made a change. So if I just pull up ChatGPT, I noticed about six days ago that the tool seemed to disappear. In other words, if you are in chatgpt when you are, when you want to use a tool at the moment, they have these three dots called View Tools. And when you click on the three dots, you have Canvas. If you specifically want to use Canvas, which we're not talking about today, but it's an amazing tool, you can then select to use Canvas. But you also had the image generation as well, and that seems to have disappeared. However, it's not a problem. It's not disappeared, it's still there. But a better way of accessing the tools when you do a chat, if you want to be specific about using a tool, is just use the forward slash command in the chat box. When you use forward slash, you'll see then it pulls up a list of all of the tools that you've got, and one of those is Create Image. And when you select that tool, that means whatever prompt you then do, it is going to create an image. It's not essential because if you in your prompt tell it you want to create an image, it will know that anyway and will probably go and create the image. But if you want to be totally sure about it, then just hit the forward slash button, select Create Image and then put your prompt. So that's how you access the image generation. It's currently inside GPT4O. I'm sure at some point they'll roll it out across other tools. Now, I said to you that Dall E3 was often ranked one of the, if not the worst, image generation tools. What's interesting is when OpenAI released this inside ChatGPT on 25th March, then in the various rankings of how these tools are performing, suddenly this went to pretty much the top of the list in a whole bunch of different areas. And that's partly because it's a very different model that they've used A lot of the past. Previous image generation tools have used diffusion models. And what OpenAI have done with this one is it's an autoregressive model and that means it does some, gives us some really interesting stuff. It has a number of really interesting benefits. And so I'm going to talk you through what's great about it. But what I'm also going to do today is I'm going to show you some practical prompts. And I like to do prompts live because I think it's more fun than ones I've pre done. The problem is the problem with and one of the downsides of this new AI image generation is it's quite slow to create the images. It's not too slow, but it's sometimes a bit frustrating if you're doing something live. So what I'm going to do is I'm going to run my first prompt right now and then I'll carry on and talk about some of the benefits and then what I will do is we'll come and have a look at what it does. So I'm just going to go back to ChatGPT. So I've already selected create an image. I did the slash and I'm just going to paste in a prompt here. And so I want to create an image of a. The concept is the trust accountant in the AI age. I've got in here the size. I want it to be widescreen. I want it to have a central figure, a confident accountant or bookkeeper standing in a modern office environment. I've got the background as a subtle blend of tech elements. I want on the left hand, left side overlay. I want a translucent checklist labeled five essential use cases with tick boxes. I want some icons. On the right hand side, I want a color palette that's maroon, white and gold accents. And I want a headline of get more out of AI and save time. And I want a sub headline of use cases for accounting firms. Okay, I'm going to come back to that. Explain. I'm going to explain some of the things that I've done in that particular prompt. But what I'm going to do right now is I'm going to send that and when you do that, you'll see that it says getting started and it's starting to respond and it takes, I think, usually about a minute. Anyway, we'll come back to that. Okay, we'll come back to that in a little while and see how it's getting on. So what is great about this new model inside ChatGPT? Well, one is the fact it's integrated already. It's inside the tool that we're already using, which makes it super convenient for us. Second thing is the conversational refinement, and I think that's partly because they've used this autoregressive model. That means that we can have a conversation like we've always done with ChatGPT, and it will not just take the prompt that you've posted, like I've just done the example, but it will actually take into account all the previous things in your conversation. So you could make the image the last thing you do as part of a conversation. For example, you could ask ChatGPT to give you some ideas on how to improve your website. You could upload an image of your website or give it the link to your website. You could ask for ideas for how to improve the website. You could have a conversation with ChatGPT about different things to improve on the website. And then when you're done with all that conversation, you could then ask ChatGPT to draw to create a design, a redesign of your, of the, of the homepage of your website. And it does a really good, cool job with that and will take into account the things you've had in the conversation before that. So I find that really powerful, this conversational refinement. The next thing that it's pretty good at is reasonably consistent character rendering. So if you've got a particular character, an image of yourself or something, it can create that and then you can have that being consistent in subsequent images. You can also upload reference images. So you can ask it to use a reference image that you upload and use that as a guide towards the style, the layout, etc. Of course, one of the things that we can do is you could upload an image of yourself and ask it to recreate the image, but with you, with a different background, for example, or you wearing a hat. Now, it's not perfect. When it then does the image, it won't look exactly like you. What I found is that some people seem to have a face that that AI can replicate better than others. In my particular case, whenever I upload an image and ask it to do something, it doesn't really look like me. It's similar, but that's going to get better over time. And there are other people I've seen do this where I thought, wow, that really does look like them. So that's something you may want to experiment with. If you want to have, for example, a practical use case might be if you want to create YouTube thumbnails and you want your image to be in the YouTube thumbnail, then you could use the tool for that. Personally, I would rather get a professional cameraman in, take some photos of me in various Poses, looks, looking shocked or happy or whatever, and using those as thumbnails. But you could use AI. I've tried it and it can be quite cool. I think that. But here's the biggest thing, the biggest thing that ChatGPT is good at is text. So one of the big problems in the past with all the tools, even things like midjourney, is they made spelling mistakes when you got text on the image. It wasn't always right. And now what ChatGPT, what a OpenAI have achieved, is getting the text extraordinarily well good. It does occasionally make mistakes, but fairly rarely. So that I think is the single most powerful thing that makes it practical for doing things like YouTube, thumbnails, social media images and so much more. The tech stuff is fantastic. And then also what we can do is we can further enhance images. And I'll show you that now because I've just noticed out the corner of my eye it did do a little while ago, it did do the image. So here is the image that I've got and the first thing you'll notice is the text is all correct. Five Essential Use Cases get more out of AI and save time Subheading Use cases for Accounting Firm owners. And that's exactly what I asked it to do in. In the prompt I wanted, on the left hand side, I wanted a translucent checklist labeled five Essential Use cases with a few tick ticked boxes. It's done exactly that. I've asked for the headline of Get More out of AI and Save time. It's done that. I've asked for a sub headline of Use cases for Accounting Firms owners and it's done that. And as far as I can see, that's perfect. And if you were to look at the rest of the prompt and compare it to what it's done, exactly what I wanted, which I think is really cool. Now of course, to get exactly what you want, a lot of the skill here, which I'll come back to later, is down to the prompt engineering. It's how you craft your prompts. You'll notice that was a fairly long prompt I did because one of the things I've been exploring is how can we craft better prompts to get exactly what we want and to get consistency in our images, which I think is really important for us as professional firm owners. If we're going to use it for things like websites, for slides, for social images, then we want things like our branding reflected. These things are important if we want to do professional images. So I think that's pretty good. But let me just talk about the further enhancement, because what we can then do is having done an image, what we can do is we can then have a further conversation. We could say to ChatGPT, yeah, that's great, but can you redo that with a different type of color? Or could we change the text? And you can absolutely do that. But even more precisely, what you can do, if you click on the image itself, as well as being able to save it, you can download the image. You will notice that you've got a little select icon and also at the bottom you've got a little prompt where it says describe what you want to add, remove or replace. So at this point you can specify now what you want it to change and it knows it's referring specifically to that image. So this is more reliable than starting another prompt later on. But even more powerful is this select button. So let's say that you wanted to put some sunglasses on him. Then on this person, what you can do is you can draw over the area and then say, turn these into sunglasses. And this just creates more clarity because it knows exactly which area of the image you want to change. It then won't change anything else. Pretty much. Sometimes there's subtle changes, but usually it's the same image. But this case with sunglasses, and I could run that and it will go and do that. It just won't happen very fast. So let me carry on. We can perhaps see what it does in a short while. So what can we do in our businesses, in our firms, how can we use this tool? What are some of the use cases that we've got? Well, some of the obvious ones are social media posts and I'll show you some examples of that in a short while. Some of the cool stuff we can do with social media posts, you can also use it for creating presentations. If you want to create, for example, templates for a slide deck or images for a slide deck, that's really useful if you do presentations. If you want to create slides, then it's a fast way of creating images. I think one of the things I love it for, and I'll show you in a little while, an example is if you, as part of your marketing, create content, let's say you've written an ebook and you want to create, you want to put in your social posts, you want to put in on your landing pages, you want an image of that ebook that makes it look like it's a real product, then you can do exactly that with this tool, as I mentioned earlier, you can do design your web pages or get it to re give you some updates on your web page. There are so many really practical things that we can do with this tool. It is absolutely extraordinary. So I think you can now see what I think is really cool about this is it's almost the same image, apart from we've gone from glasses to sunglasses now it has made some subtle changes. As I said, the face is just a little bit different, but probably because where I drew the little thing changed the face as well when I did the select tool. But if you look at it, the text is the same, the logo's the same, the background's 99% the same. So that ability to select just one little piece of your image and get it to just change that and leave the rest, I think is really powerful if you start to fine tune your images. So let me run another prompt and I just need to find it so I can copy and paste it so that you don't, don't watch me type this because it's a long prompt. I'm just going to grab the prompt control C. I'm going to go into the box, I will. To make sure it doesn't image, I will hit the forward slash and choose Create Image. And you'll see that appears in blue. And then I'll paste in my prompt, which is, I'm going to show you this prompt in a short while. We'll go through this prompt. It's a prompt to create a social media image. Let me just run that prompt and we're going to come back to that a short while because what I'm going to do is show you some other types of images that you can create. So I mentioned earlier, I mentioned earlier that if you've written an ebook, for example, or you've created some sort of digital product, it could be an ebook, it could be perhaps you've got an audio program that you've done and when you put it on social images, your website, your landing pages, you want to make it look like a real product. And so one of the things that you will often see is graphic designers turning things into 3D images. And so I know this is quite small on the screen, but one of the things that I created some years ago was I did a benchmarking report for the accounting profession, which was all about. I'd researched and benchmarked accounting firms on how much they price, how much they price for different services. And I created a detailed research report around that. And I at the time turned that into a product and My graphic designer designed it and made it a 3D image which you can see on the screen, that little 3D image. And then what I also did very recently is I wrote a 35 page research report for the accounting profession on privacy and security, on client confidentiality. And I decided I would like to have that done as a 3D image. So what I did was I created a fairly simple little prompt that says I have attached a 3D product shot and the front cover of my report on pricing and security. I want you to turn the front cover of my report into a 3D product shot simple prompt. And it didn't need to be complex because as I said earlier, one of the great things about this tool is that we can upload images to act as references. So in this case I just took 3D image of a product. It happened to be one that an existing product I've got. But you can get these images from the website if you find a particular 3D style that you like. I just uploaded the PNG file and then I also uploaded the front cover of my privacy and security book guide. I think it was a png. I think I did a screenshot of the PDF. It was probably as simple as that. And I uploaded that and I ran that prompt and what it came up with was this image on the screen which I think you'll agree is almost exactly the same style as the reference image. The three, the other 3D image, but this time it's my new guide. And what I think is really powerful about this is again the text is absolutely accurate, no spelling mistakes. It's got the logo right as well. And it's also got the logo in obviously recognize that it's not square on. So it's got the perspective of the text and so on. Really done really well. It looks like a. It looks like it was what you would have done with a graphic designer doing it or with software used. You can still buy to do 3D images, but now we can do it here. So I think that's a really powerful use case for accounting firms if you're doing content marketing, written an ebook, an audio program, other digital products, very easy to create a nice little 3D image to go on your website and so on. Okay, what else can we use it for? Well, social media posts are great. So if, for example, one of the things that you could do if you have a Facebook group or a Facebook page or other social pages, you might want to post, for example once a week a motivational quote or them the week and you can create some really cool images doing that. What else could you create? One of the things that you can do is it's really good at creating photorealistic images. For example, if you're creating a brochure for your firm or you want images for your website, then now you can create them really quickly inside ChatGPT and nobody will really tell that they're AI generated. Now my view is that when you have a website and you have images on your website. I've always hated stock footage. I think that when you do a website, the images should be of you and your team because people buy from people. I'm not a big fan of stock images, but there is a place where you might want to use them. But what I did want to show you about this particular image and one of the things it can do really well is when I created the prompt for this image, I also uploaded the logo for the Value Pricing Academy. I asked it to put it in the image. I kept it really vague. I didn't say where to put it. I just said include it somewhere. And as you can see, it's put it on the wall. And what I like about this is that again, it's got the perspective pretty much right. It looks like that is the logo painted onto the wall. There's no spelling mistakes. You'd be hard pressed to know that was AI generated. So one of the things that you might want to do when you're doing images for your accounting firm is upload your logo. It's not perfect every single time. I think I've been lucky. That was one shot did it right the very first time. So that's, I think pretty cool. The fact that we can include our logos, we include our color schemes in images. Here's just another example of a image I did which again, you might want to use this in a brochure in on a website subject to my suggestion that I personally don't like. I don't like anything other than proper personalized images of your team and so on. But nevertheless, you might find use cases for images like this in brochures and so on. Another thing I mentioned is you can create use it for creating slide or slide backgrounds. So this, for example, was I go to create me a slide background that I could use. I won't use this. I was doing this for demo purposes. But one of the things it knows is that when you create slides backgrounds, you don't want it to be bold colors because you want to be fairly muted because you don't want it to distract from the messaging that you put on it. So this I thought was pretty good. And then what you can do is if you want images on your slides, you can then get images created like for example, this image. And one of the things that, if you're creating images to go on top of your slides, one of the things that you can do which is really cool, is you can specify that the image needs to have transparent background, which basically means that this particular image, I can put it onto the slide and it'll, it's got the slide coming, the slide image coming through. I could put it anywhere I want. Whereas if it was a plain white background, you'd have a big white box sat on top of your slide background. So that's pretty cool. Okay, what's, where do we get to? Yes, let me show you. Let me show you where we got to with one of our images. So I put in, if I go back to chat GPT, I put in this particular prompt, this prompt here. And I'll explain it in a second. Let me just quickly show you a sneak preview. That's what it came up with just now. And so the idea here was I was thinking, okay, well what if we do want to do social media posts for a motivational quote of the week? Let's say if you want to, if you want to build, if you want to do some more engaging social posts with quotes, then this is the prompt that I use. I want a minimalist top down image of a thoughtfully arranged workspace designed to inspire calm folks and quiet productivity. I want it to be square. I want it to be a top down bird's eye view, clean and balanced layout with intentional white space. I want soft neutral tones, for example, warm whites, pale beige, gentle morning sunlight casting natural shadows, evoking a peaceful and provocative mood. Then I want some specific key things on the desk. I want an hourglass to suggest the passage of time. I want a travel brochure with the front cover showing a paradise beach location in the Maldives. And I want an iPad showing a goal planner with an Apple pencil including sketches suggesting the planning of goals. Now, I chose those three things in this prompt because I was using this particular quote, this motivational quote. It's Never Too Late to make a Change by Paolo Coelho. Each object should be spaced thoughtfully with visual breathing room. The mood wants to be calm, focused and subtly uplifting, designed to stimulate, simulate the start of a mindful and productive morning. I've then given instructions on how I want it to do the prompt. Sorry, the quote I want it to be deep gray or soft charcoal centered text with generous line spacing. And I've also specified the use case I'm going to use this for. I've told it I want to use it for social media posts, productivity blogs or morning routine campaigns. And so this is what's created and I've only seen it. You saw me doing it live. I think this is really good. It's exactly what I wanted. And so if you've got a motivational quote, you can see it's never too late to make a change. It's got the text perfectly right, exactly how I wanted it. It's got the three images, which was an iPad pencil with some images around setting goals, a travel brochure to the Maldives to a beach, and an hourglass to represent the passing of time because it's never too late to make a change. So, and here's the really important point I want to make, is that when you understand how to construct your prompts and you've got some prompt engineering skills and you can start creating some templates, you can start to create consistency in your outputs. So let's say, for example, let's say you like this particular prompt or this image. So if you like this image and let's say you did want to post every single Monday a quote of the quote of the week, then because of your branding and style, you'll probably want every single week to have some sort of theme commonality with your images. So think about your branding, think about having some consistency in your marketing, your social posts. And so what I did with this particular one is I, first of all, I got the quote I actually used, I don't know if you've used it. It may be a session for another day. I have already built into ChatGPT, I'm using tasks, so I get every single week it gives me automatically a motivational quote and I get a notification to tell me when I've got a quote. And then what I did is I ran a prompt and said based on this quote, what would be the three suitable images? And ChatGPT said an hourglass to represent the passing of time, a brochure, an iPad with some scribbles around goals. So then all you would do is you'd have a template prompt that gives you the style of image you want and then all you have to do is put in the specific quote and what are the three objects and then you get up, get your image. You could then get somebody in your team every single week just to run tasks to Get a quote every week. You could get them to run a standard prompt every week to then get three images, sorry, three objects to have in there. And then you could run your standard, they could run your standard prompt, which will then give you an image that's going to have this consistency with the particular design layout that you want, the colors you want, and so on. So I'm not saying necessarily use that particular style of image. I personally like that. It's clean, it's simple. There's a lot of things I do like about that. But the key point here is if you want to get that consistency in your prompts, if you've got particular styles branding for your firm, then having some templates for the way that you do your prompts is really powerful. So let me just share with you a few tips around AI image generation. And the first is that as with all things with AI, with large language models, with ChatGPT, if you want to get better results out of the tool, then you need to understand how to craft your prompts. This is the world of prompt engineering. And I know many people know that. I've been teaching for the last two years now how to do prompt engineering. And it's just the same with images, it's just a different style of prompt. If you want to get the best results from your images, you need to know how to craft prompts. And I, about a year and a half ago, I started playing with midjourney. I bought some courses on how you craft prompts in midjourney. So when this tool came out, the first thing I did was went back to all my notes of how to craft prompts in midjourney, and then just test them out to see to what extent they have to be changed for this particular tool with ChatGPT, and so when we're creating our prompts, if you want to get some consistency in your outcomes, then there are a number of things that we need to think about and specify in our prompts. For example, you need to understand what are the different types of image styles and what image style are you looking for? There are dozens and dozens of styles, for example, and different styles for different purposes. So for example, if you are doing a hero shot on your website, an image on in a brochure, for example, then you might want a photorealistic image that's a particular style and one that I tend to use quite a lot. But even with a photorealistic image, there are things that we can do that we can specify to get a particular type of photorealistic Image. And this helps if you understand photography, have some skills around different types of film, lenses and so on. Because when you understand that, we can be very specific in our prompts as to exactly what type of photorealistic image. But there's more than just photorealistic. We can have things like flat images, we can have cartoon style images. There are so many different cartoon styles, Lego and many more. So there are lots of different styles of image that we can create. You need to understand that because then you can specify in your prompts what is the most appropriate style style for the thing that you are looking for. Sometimes it might be photorealistic, but sometimes it's not. If you're looking to create, for example, a social post, perhaps you want to create an infographic, then a different style will be more appropriate. You also need to understand aspect ratios. So that's the size of the image, if you like. If it's a PowerPoint presentation, that would be 16 by 9. If it's a social image, very often they are small square. And so if you want to have some control over the size of the image, then put the aspect ratio in there. Now, unfortunately, one of the things I'm finding is that it doesn't always follow exactly. You can't really fine tune exactly what the aspect ratio is. I think midjourney is better for that. Midjourney. You can be very specific in your prompt. If you say I think it's a curly symbol and then AR for actor act, aspect ratio, colon, sorry, aspect ratio. And then let's say rather than 16 by 9, you want 16 by 7 because you want it not to fill your slide up. It will do exactly that in mid journey in ChatGPT at the moment it approximates it. And so you might want to simplify it by just saying widescreen square portrait, for example, I'm finding that if I say 16 by 9, it will do that. But if I want a 16 by 8, it doesn't do that. Exactly. Nevertheless, it'll get better. It'll get better. No doubt you still have an element of control. Just make sure your prompt is specific. You also want to be specific about things like the text that you want. Think about other things to specify. And I showed some examples earlier, things like the color palette, things like the mood that you're looking for in the image can all impact on the image that you are getting. So you need to understand how to create images. Image prompts, you need to understand if you want to get the best out of it. All the different elements to build into your prompting and then what you want to do is to create consistency. I strongly recommend you start building prompt templates. So you will probably once you start to realize that you might want to have a particular style for your social media posts, you might want to have a different style for your website. For example, you might want to have a different one for PowerPoint images. Then start to build up a library of prompt templates where like the one I showed you earlier where if it's a motivational quote of the week, you have the template and then all you do is change some specifics like the objects or the text. But the rest of it, it'll keep that style, it'll keep that color theme and so on. Okay, let me just show you one more final example and then we will wrap up. So I did this one just a few days ago and I used a prompt for this particular one. And the reason I did this is because I did a course, I did a course A week ago, 55 minute video training for accounting firm owners all about privacy and security. Because this is a big issue, I think that probably, I suspect almost most, I'm going to guess 90% of accounting firm owners using Chat GPT are probably unwittingly falling foul of professional guidelines and breaking the data protection rules. We have to be really careful about what we do put into these AI models, particularly accounting firm owners, because of client confidentiality. So I did this course to talk about what the problems are, but also the strategies and tips on how to make sure we don't fall foul of the professional requirements and don't fall foul of data protection laws. So that's my image that I asked it to come up with at the end of my conversation around I gave it the transcripts, I got it to summarize all the marketing material and then at the end of it I then put a post saying create a suitable image for me. I gave it a template prompt and this was first time around it created this image which I'm really pleased with. It got the text perfect exactly as I wanted it. So I love this tool. I think it's extraordinary. And bearing in mind as I'm recording this, it's only been out for about four weeks. This is just going to get better and better. This is the start of this journey with this new autoregressive tool that OpenAI have released and it's just got so many practical use cases as well as of course the fun stuff. Thank you so much everyone for coming along and giving up your precious time. I hope you found this valuable. Bye for now.