317 | Stop Creating AI Slop: How to Build High-Converting Visual Content with AI with Aastha Taneja

Summary

---

Leveraging AI – Episode 317: “Stop Creating AI Slop: How to Build High-Converting Visual Content with AI”

Host: Isar Meitis
Guest: Aastha Taneja
Date: August 11, 2026

---

Episode Overview

This episode of Leveraging AI centers on moving beyond “AI slop”—generic, low-engagement digital assets churned out by AI tools—toward creating high-converting, visually striking content using AI. Host Isar Meitis and creative expert Aastha Taneja unravel practical, step-by-step strategies to build impactful digital visuals for websites, social platforms, and eCommerce, harnessing both traditional design principles and cutting-edge AI workflows.

---

Key Discussion Points & Insights

What Differentiates High-Quality from Generic AI Content

• AI Slop Defined:
  Content made with minimal effort, relying entirely on basic prompts or default settings, resulting in bland, widely-recognizable AI-generated design.

  “AI slop is basically when you are not actually using AI smartly… you are just typing big prompt and asking ChatGPT or any LLM to create the best output for you.”
  —Aastha Taneja [04:47]

• Combining Human Creativity with AI Superpowers:
  Success comes from integrating foundational design skills and brand context with AI’s capabilities. Full dependence on AI produces average results; human refinement is essential.

  “If you want the perfect output, you have to integrate traditional method and AI method.”
  —Aastha Taneja [05:33]

Preparation Before Content Creation (Inputs Matter!)

• Start with Mood Boards:
  Most skip this step. Gather and curate visual references (images, color palettes, font styles) before prompting AI.

  “The starting primary foundation is creating mood boards… if you’re not clear what you want, AI cannot help you.”
  —Aastha Taneja [06:53]

• Analyze Top-Performing Industry Content:
  Emulate engagement patterns from market leaders rather than copying designs; extract formats and hooks, not just visuals.

  “If a specific brand…are consistently driving a lot of engagement, it’s because they’re doing something right… How can I use AI to learn from that?”
  —Isar Meitis [08:17]

Tools & Workflow Deep Dive

• Mood Board Preparation:
  Use Figma (free version suffices) or Pinterest. Collate and screenshot aesthetic references ([09:25]).

• Project Feature in LLMs:
  Set up context universes (“projects”) in ChatGPT or Claude. Feed them brand websites and assets so the AI understands the full business/visual context ([11:01]).

• Training AI on Aesthetics:
  Provide references not to copy but to teach AI nuances (typography, color, formats).

  “Not to copy and paste, but basically training my AI on what kind of aesthetics I am going to…generate.”
  —Aastha Taneja [13:18]

• Leveraging Viral Patterns:
  Learn from viral content formats. Focus on replicating the formula (structure, hooks, style) rather than the visual identity ([15:13]).

• Automating Analysis of Engaging Content:
  Asta and Isar recommend and practice creating recurring AI routines (content “machines,” e.g., for LinkedIn or YouTube) that constantly analyze top performers, extracting trends and engagement hooks ([17:32]).

Practical Execution: From Ideation to AI Refinement

• Iterative Refinement Loop:

  1. Generate initial drafts in ChatGPT using references and context.
  2. If product consistency/complexity challenges arise (common with AI), pass output to “Nano Banana Pro,” Google Gemini, or Magnific VV for enhancement.
  3. For precise fixes, manually erase incorrect sections (using Photoshop, Canva, or even PowerPoint), then use AI to fill gaps with the correct details, providing both the full product and focused close-ups for reference ([24:30]).

  “If you can make it focus on the three things you wanted to get right, which is exactly what ASTA is doing in this process, it will get it right.”
  —Isar Meitis [25:55]

Advanced Automation and Scaling

• Building Node-Based Workflows:
  Tools like Magnific, Weavy, and Preterm allow for node-based flowchart processes—set up workflows once, reuse for different projects by swapping reference inputs ([29:33], [31:06]).

  • Preterm’s AI “brain” can analyze a brand’s website and references, then auto-generate batch workflow prompts for mass asset generation ([31:06]).

• Claude Skills & SOP Extraction:
  Leverage top YouTube tutorials or industry processes: transcribe, upload to Claude, and convert into reusable Skills (SOP MD files), enabling rapid, repeatable execution at professional standards.

  “Claude skills is something people have not been using…After you are done with that, just add everything to Claude and ask Claude to create a skill file for you.”
  —Aastha Taneja [35:14]

---

Notable Quotes & Memorable Moments

• On Preparation:

  “Major thing is you are not creating a mood, but you are yourself not clear what you want. So you have to be cautious about what exactly do you have in mind. AI is there to just be your assistant.”
  —Aastha Taneja [07:17]

• On AI as an Amplifier, Not a Replacement:

  “Everybody thinks AI will replace your job. I believe that if you have a strong foundational background, you can use AI way too effectively.”
  —Aastha Taneja [07:48]

• On Reusing Viral Formulas:

  “The chances of getting viral are much more than creating something for completely scratch…The formatting are completely different from the visual aesthetics.”
  —Aastha Taneja [14:55]

• On Visual Consistency:

  “They look fantastic. And they look consistent. Because ASTA took the time to train AI on how to create this kind of visuals.”
  —Isar Meitis [21:16]

• On Automation at Scale:

  “If you have these kinds of systems in the back end, again with efficient designers who can…add 30% of refinement to this, you can create a bunch of assets in a very short amount of time.”
  —Aastha Taneja [34:02]

---

Timestamps for Important Segments

• 00:00 – Introduction and Episode Theme
• 03:38 – What Creates “AI Slop” vs. Distinctive Content
• 06:50 – Essential Preparation Steps Before Prompting AI
• 09:16 – Mood Boards: Tools, Creation Process & Best Practices
• 11:01 – Utilizing Projects in ChatGPT and Claude
• 13:18 – Training AI using Aesthetic and Typographic References
• 15:13 – Leveraging Human Psychology & Viral Patterns
• 17:32 – Automated Competitive Analysis Using AI
• 21:16 – Visual Consistency: Examples and Results
• 24:30 – Practical Tip: Overcoming AI Errors via Layered Workflow
• 29:33 – Node-Based Automation: Scaling with Tools
• 31:06 – Preterm: AI Brain Automation Example
• 35:14 – Claude Skills and Automating Prompts/SOPs
• 37:27 – Recap and Closing Insights

---

Episode Recap

• Core Message:
  Simply prompting AI yields bland, interchangeable visuals (“AI slop”). Instead, infuse the process with creative context—brand DNA, mood boards, references, viral formulas, and human refinement—to unlock stunning, conversion-driving designs.
• Toolstack Highlighted:
  Figma, Pinterest, ChatGPT, Claude, Magnific/Weavy/Preterm (node-based automation), Gemini, Nano Banana Pro, Claude Skills, NodeGPT for transcripts/SOPs.
• Scalable Workflows:
  Use platforms that let you template processes, batch-produce content, and leverage best-in-class SOPs—while always refining 30–40% of output for uniqueness and brand fit.

---

Where to Find Aastha Taneja

• LinkedIn: Main channel for networking and insights
• Instagram: Recently started, focused on practical content tips

“If people wants to reach out to me, they can always just text me on LinkedIn. That is the best platform.”
—Aastha Taneja [38:56]

---

Final Thoughts

This episode provides a practical, actionable roadmap for anyone looking to move past copycat “AI slop” content. Key is to combine deep brand/contextual understanding, leverage reference material and viral content analysis, and refine outputs—using both traditional and AI tools. The approaches shared let even non-designers create assets that rival top creative agencies in effectiveness and style.

---

For visuals and workflow demonstrations referenced in this episode, visit the Leveraging AI YouTube channel (as mentioned by Isar).

Transcript

[A]

Hello and welcome to Leveraging AI, the podcast that shares practical, ethical ways to leverage AI to improve efficiency, grow your business and advance your career. This is Isar Matis, your host, and I've got a really, really amazing episode for you today. Every business regarder, whether it's large or small, regardless which industry it's in, creates digital marketing assets. These may be videos, these may be images, these may be for social media, for your website, for online sales platforms like Amazon, Shopify and so on. We create visual assets all the time. The problem is most companies that are using AI to create assets and everybody's using AI to create assets. The problem is that most people are doing it in a very superficial level and hence they're getting generic results that look like what everybody else is generating, which means it looks cool, but it drives exactly zero engagement or very low engagement and, and very low business results. And we're doing all of this to drive business results. So what we're going to do in today's episodes, we're going to learn a step by step detailed process on how to create digital assets with AI that actually move the needle. Things that when you put them on your website, on social, on storefronts, Amazon, etc. You are going to get engagement and you're going to drive clicks and you're going to drive people to your websites or to wherever is that you're trying to drive them and, and at the end of the day, make business results obvious versus oh, I'm just putting content out there because I know I need to do this. Our guest today, Astara Tunija, has been a creative and in the design space for a while. She worked before AI building digital assets for brands as well as doing a lot of freelance work. So she really understands what's required from a good design in order to actually work, because she's done this for a living for, for more than a few years. So she in the past few years has taken what she knows as solid, good digital assets and she taught herself and as she's teaching others how to do this effectively with AI. What are the steps, which tools to use? What are the little tricks that nobody knows how to create really amazing skills inside? Not skills for yourself, AI skills that do what the top creatives in your area does without actually knowing how to do that. All of these little amazing tricks that at the end of the day allows you to use AI to effectively create real digital assets that actually move the needle, which is something that, as I mentioned, every single company and every single organization and individuals if you're a solopreneur, everybody needs that because we're all depending on this in order to drive traffic to the different assets of our business. In addition, in our short conversation that we had preparing for this, she gave me gazillion practical tips which you're going to get today. Lots of really amazing tricks that you need to know in order to make this better without a lot of effort, which really, really is amazing. So I'm really excited to welcome Asta to the show. Asta, welcome to Leveraging AI.

[B]

Thank you so much for having me. I'm so excited for this conversation. I really wanted to keep this very raw. I'm not prepared anything, so I really want to share some practical tips with the audience and I hope it helps them practically so for sure.

[A]

But before I. So you know, even, even in a little conversation that we had last week preparing for this, I'm like, holy crap, like there's so many cool little things that people can do. And so, but, but let's start really at the, at the high level and then we're going to dive immediately into the practical stuff. What is the main things that differentiate good visual content from blah vanilla generic v visual content?

[B]

Honestly, it depends upon a lot of factors, especially with AI. I feel that majorly people are using CHARGE GPT, okay? You can see a design and you can visually tell it that this is being made by ChatGPT. I'm going to show you tricks and trips. I use CHAR GPT almost every single day. I use Claude every single day. But the level of output you can get using chargpt is very different from what people are generating. Again, not criticizing anybody, but you have to be very mindful while using AI. You have to be very smart. Everybody has access to AI tools, every single person out there, okay? You have to be very mindful, you have to be very smart on which model to use for which practical output, which, you know, which conversation, what kind of conversation you have to do with charge people are still dependent upon writing big ass prompts and you know, creating those big paragraphs. And okay, this is what, this is how they will get the perfect output. And that's what I see, you know, usually on screen that people are getting very, you must have heard that word AI slop. So that AI slop is basically when you are not actually using AI smartly, you are not, you know, using your brain. You are just typing big prompt and asking chargpt or any LLM to create the best output for you. That's what I believe that people have been doing. What differentiates from a good content from a blah content is majorly that if you're not, you know, being creative, you are just totally dependent upon AI, which I believe should never be happening. I know everybody has access to AI and everybody's like, okay, now we are going to say bye bye to our traditional methods and we have to be fully dependent upon AI to generate everything from A to Z. Because I do a lot of corporate trainings and whatever I see people currently are doing is they are like, okay, we have AI. I have dumped everything to AI. Now AI is not generating the perfect output means this model is wrong. Or this like the, the way it works is not, not right. But the thing is you have, if you want the perfect output, you have to integrate traditional method and AI method. You can't be fully dependent upon AI, at least in this stage of time for sure. Maybe eventually we will reach a point where we can say, okay, 90% things are done by AI and just 10% is manual. But I believe in today's era we are at least 60% there. 60% can be done with AI, but 40% with good refinement is very important. And AI slop is majorly. When you are just dependent upon AI creating the one output and posting it and using it for your businesses, that is AI slop. So you have to make sure whatever you are creating with AI, it needs to have certain kind of aesthetics to it and not just an AI slop. So you have to be mindful for the same. And I'll share that. Tips and tricks on how to do that.

[A]

Yeah, so let's really dive in. And I think the first thing that we talked about previously is, is what do you do before asking AI to produce the content? Like what are the inputs that make a difference in order to create the output as a better output? And I think that's what a lot of people miss. Like you're saying a lot of people jump in and they either copied a prompt from somewhere because somebody said this is an amazing prompt to do something and then they add their two cents to it. So what are the, if you want like the preparation steps or the things that needs to be as part of the input in the process before you ask for the output.

[B]

Okay. So what I usually do is, and most of the big companies that I've worked with, their major step, the starting primary foundation is creating mood boards. That's, that's what I think is the most important step. Which people, most of the people miss what they do Is they just again, as I mentioned, go to ChatGPT, just say that, okay, this is the design that I'm looking for. And they try to create kind of a prompt, or maybe ask charge to write a prompt based on their personal input. And then they feel that, okay, the result is not that good. Major thing is you are not creating a mood, but you are yourself not clear what you want. So you have to be quite, you know, cautious about what exactly do you have in mind? AI is there to just be your assistant. You cannot be fully dependent upon AI. You have to have a creative mind. That's why I always say that people, because most of the people think that AI will replace your job. I believe that if you have a background, if you have a strong foundational background, you can use AI way too effectively. Then comparison in comparative to a person who doesn't have that creative background. For example, if you're talking about design and visual aesthetics, a person who have had spent a lot of years in creating those designs themselves using their hands, or by hands I mean using Photoshop, Illustrator and all the traditional platforms out there. If you have done that, if you have done that hard work, the way you will use AI will be very different from the person who is just fully dependent upon AI.

[A]

So I agree with you 100%. But I will add something else that, that we also talked about and I do in many different, in many different ways myself. And you I know do it as well because we talked about this. As much as you don't have knowledge in all the fields, like I've never written code, I'm not a designer and all these things because I exist today, you can relatively easy copy the people who are doing this in the best way. And how do you know they're doing it in the best way? You, you don't know how to tell good design from bad design, but you can easily see what people are engaging with. So if a specific brand or a specific person in your field are consistently driving a lot of engagement, it's because they're doing something right. So if I see something in my field that is doing something right, how can I use AI to learn from that, to produce similar kind of outputs? And you can totally combine the knowledge, the mood boards into that. Because I think most people don't even know what mood boards are. They definitely not how to produce mood boards. But I think combining these two things together will help a lot.

[B]

Yeah, so how I usually do is I use Figma for my mood board stuff. Figma and Pinterest Pinterest is one of the best platforms out there. So for example, if I have to create some visual assets for l', Oreal, this is one of the famous brands that sells shampoos. Okay. For example, if I have to create visual assets for them, I usually start by going to Pinterest or any website that you feel that you get most inspired from. I take all the visual aesthetics that I personally really like. Okay? Obviously it doesn't mean that you have to copy and paste each and every design, it just means that whatever, like what are your brand aesthetics? Like what do you have in mind before, you know, even, you know, asking AI to, you know, help you with the assets? So what I usually do is I download all the best looking assets that I personally like. I just drop them to Figma, I take a screenshot and I then I put that to ChatGPT. But before that I want to really clarify one point. People feel that Claude is best for everything, which I completely agree with. But people might disagree with me on this one point. That charge has its own capabilities and people have not been using ChatGPT effectively. Claude is great for complex, complex tasks, especially when you're trying to build apps. And when I say apps, you don't have to be a coder, you don't have to be have a very technical background for that. I don't have any technical background, but I have built apps for different companies, custom solutions for them. So I, for design, for creativity purposes, I especially am dependent upon ChatGPT. It gives perfect output, trust me on that. So how I usually do is after creating the mood board, I go to ChatGPT create a project. Most of the people are not using the project feature, either it is in Claude or ChatGPT create a project. If you want me like, I'll just show that to you.

[A]

Let me just so two things while you're opening and sharing your screen. One, for those of you who don't know what a mood board is, a mood board is a bunch of images that are glued together with color palette and specific styles. And if it's objects, then specific objects. If it's an atmosphere, then specific images of that atmosphere that really illustrate the kind of style you're looking for. A few weeks ago I've done a whole workshop for interior designers and people from that field, so architects and so on, showing them how you can take images of styles that your client likes, drop them into an AI automated process and it creates the mood board for you. And the mood board knows how to. And the AI like I said ChatGPT is really good today, really, really good at combining together multiple standalone images into a mood board. And you can ask for four different versions and find the one you like the most. So this is about the mood board, about projects, which ASTA is going to show us in a second. Projects are a feature inside of ChatGPT and inside of Claude that lives on the left side menu and, and it allows you to create a, I like to call them context universes. You can create a little context universe, a little bubble of context that has a lot more information about a specific topic that the rest of CLAUDE or The rest of ChatGPT doesn't. As long as you work within that bubble, then it knows a lot more stuff about that topic. In this particular case, specific design, specific aspects that ASTA is going to explain in a second.

[B]

Yeah, so for example, if I am working with the brand that sells shampoo and I want to create a mood board for that brand, whatever images I like, I'll just copy them directly. Open Figma. You don't need a paid version of figma. If you, if you can work around with the free version, it's best. So I'll just copy all the images that I really like. I'm just randomly copying anything as of now. So like you can basically add tons of images here, this is your board. Once you do that, just take a screenshot. Okay.

[A]

Of all the images together.

[B]

Yeah, all the images together. That's, that's what I do. It's just easier to even, you know, paste, Copy and paste 1, 1 image. All you can do is just add all the images and take a screenshot. Okay, but this is the after step, I'll, I'll show you how I start. So I'll create a project. If you see on the left hand side there's a project option, right? Yeah, you click on new and just say l', Oreal, for example. Now after creating the project, I'll go to my website, for example, my client's website. I'll basically add the URL and just say study. That's it, just study about this. Like obviously you can, you can enhance it more just for the sake purpose. I'm just trying to say that all you need to do is just add your website, add whatever information you have for the brand, just let LLM study about it. You can use Claude as well for this, whatever you like. But chargeability works pretty well for me, at least for this stage. So I won't, you know, spend more credits by using Claude. So you have to be mindful on how many credits you are wasting. So for for easier tasks, I think charge you is perfect. So not only the website, drop all the assets, for example you have SOPS for your brand, for example, just drop it to ChatGPT. Add whatever information you have. After you are done with this, just take a screenshot and just paste it to ChatGPT in this chart. Only once you create a project. Now ChatGPT knows about this brand. It knows everything. Like how, how many categories of product it has, what are the color palettes, each and everything. Just throw in your mood board and ask. Okay, I want to create visual assets for this brand. I want to create assets for Instagram, TikTok for different platforms, maybe for, for. For different kinds of platform, maybe for E commerce website for Amazon listing images, main images, A plus content, like anything. Just describe what you are trying to make. Okay, once you are done with that, ChatGPT is going to analyze each and everything. Even one more thing. Most of the brands miss that. For example, I really like how the text is written in this. Okay, I'll take, I'll just copy this. I'll again paste it to ChatGPT. I'll ask you to understand the visual aesthetic of how the typography works. Again the rule or the main conclusion is not to copy and paste, but basically training my AI on what kind of aesthetics I am going to, you know, I want to generate basically and that is how once you have like put in all your information to charge, then ask chargeability. Okay, I want to create visual assets. Now tell me what do you think would go viral? Now chargeability again will give its own, you know, versions. Okay, this, this, this. But you again you can't be dependent upon that. One thing I really want to share here, human psychology. Because I've worked in many brands that really wants to create their products viral and all of that stuff. So we do that on every single day basis. So if you really want to create something that has the virality aspect, though I believe if you're providing good value to your customers, virality is not something you should look forward to. But some brands or some clients are very reluctant to that. So you should always keep in mind whatever patterns that your human mind is likely to engage to the kind of content they have seen before. So this is a psychology and this is like there's a big research happened on this. Like whatever my brain has seen in the past that I have engaged with like the kind of content that has gone viral. If I try to recreate for My brand using that same format, it is bound to get viral. Like the chances of getting viral are much more than creating something for completely scratch. Now people will disagree and say that, okay, it means why we have to just copy and paste. No, it's not about copy and pasting, it's about taking the viral format. Again, people misunderstand this and say that, okay, if I'm copying a format, it means I have to copy each and everything. The way the visual languages or the way the aesthetics look. No, the formatting are completely different from the visual aesthetics. You have to understand how the hooks are, what are the lines that they are seeing before starting their content, what are the visual aesthetics, the kind of elements they are using. Just try to ask Chargpt that this is something that I really, really like. How, how can I do this for my brand? Then the output that Charge will give you will be far better than just being fully dependent upon it and saying that, okay, create some, like, give me prompts to create assets. That doesn't work, you know, beautifully. So you.

[A]

I love that. I want to, I want to summarize a few critical points and add my 2 cents. So the, the inputs that you're giving here are not just, oh, I want to create this and that. The inputs included work that you had to do on several different aspects. Aspect number one is understanding the brand, like from their website, from their previous work, from whatever it is that you want. It's like, okay, go and learn this brand. So now it understands the brand. In many cases you will do the same for the target audience. Who is their target audience? What do we know about them? So that becomes another input to this. The third component is really the visual look and feel. And as you mentioned, just as a summary, there's different aspects to the look and feel. Some of it is the overall aesthetics. Like, I just like this kind of style. Some of it could be a font, I really love this font or the way the font is laid out on the page. So show it examples and ask it to learn that and so on. But then there's a little pixie dust, like a little magic that that makes things work for others in your industry. In my. So I have built, I call it my content machine. So I have a LinkedIn content machine, a YouTube content machine and so on. And what these content machines do as the first step in all of them, they go once a week and they go and look for people in my industry, in my case AI and so on. But you can do this for any industry and it finds the best performers on those platforms for that week. What post got the most amount of engagement in that week? And then AI picks 20, 30, 50 of those and it says, okay, what do they have in common? Like, what do they have in common? Are they all videos? Are they all still images? If it's a video, how long is the video? What's in the video? Is it a person? Is it text? Is it a combination of text? And like, it really tries to analyze every one of these components. How long is the text? What was the hook? What are the gaps between the lines in the text? Like, it tries to analyze all of that. Now, I know it sounds like a lot of work. I do zero of that work. I've got a set of skills that I've built that do this on regular basis at scale. And then it analyzes and says, okay, what's working right now on LinkedIn is this. And then going back to what Asta said, I'm not copying what they're doing. I'm learning the format, the aesthetics, the delivery method. And then I'm taking my content, the things that I want to say, and it merges the two together. So I'm using a. If you want a successful formula, proven formula, together with the things that make me and my content, me versus generic slop, and then the both of them together work very well. And you just describe the process on how to do this very easily.

[B]

Yeah. And one more thing I want to add to this, I just want to show the difference. So this is called an AI slop. So if you simply ask ChatGPT to create a poster for your brand, this is how usually the results are without giving any references, like these are AI slots. When you're like, you can see and just tell that okay, this is being made by AI. But when you give good references, when you try to make AI understand what kind of visual language are you trying to create, you get these kinds of results. Most of the luxury brands, they go with very minimalistic look. They have certain kind of fonts they want to try with. Now for this particular poster, I just screenshotted few of the fonts that I really, really liked. I didn't want to generate anything from scratch, so I gave it a few references that this is the kind of vibe I'm looking for, some minimalism. This is the font that I'm looking for. I took screenshots from way to different things, created a mood board, added to charge GPT and this was the result. If you see the kind of fonts it has used. It's very different from usually what ChatGPT generates. It's very different from that. So that is the power of adding references and adding some mood boards to any LLM that you are using.

[A]

For those of you who are not watching the screen and just listening to this, I'll try to describe what we're seeing. So it's a bunch of ads that look extremely luxury. That's the feel that I get from these. And they're very minimalistic and they have consistent style. So like the style of them are very much the same and it is very different than, again as that said that like, okay, this is what ChatGPT or, or other things are doing. They look extremely professional, very, very clean. But they all look the same. What I mean by the same. They're very different, but they all have the same very clear style, the same kind of font, the same kind of minimalism, the same kind of brownish grayish kind of like background. And they just look fantastic. And they look fantastic. And they look consistent. Because ASTA took the time to train AI on how to create this kind of visuals. Now people think, well, I don't know how to train AI. I don't even know what that means. Well, we just told you, all you need is a ChatGPT project with the right assets, with the right background, with a little bit of instructions and AI can learn the style that you want and then it can reproduce the same outputs or the same style of outputs when you give it new inputs.

[B]

Yeah, yeah. And one more thing I want to add, for example, some of the products are very complex than others. ChatGPT is perfect for design. Okay? You can get really good design aesthetics from it. Nano Banana Pro, which people still use a lot, which is great, but it's majorly for consistency. That's what the way I usually do it, I just train ChatGPT around my product, around the visual style that I'm looking for. I asked GPT to create this poster for me, okay, any poster. And then I feel that, okay, my product is not looking right because when you are posting on social media, you have to make sure your product needs to look exactly the same. So how do you do that? ChatGPT still struggles when it comes to product consistency or, you know, sometimes it struggles. So once you are done generating the visual aesthetic, you just upload the same image to nanobanana Pro. You can use any platform. Either you can use Google Gemini directly or you can use Magnific VV or any other. There are a bunch of platforms out There. So just ask ChatGPT that I want to recreate the same ad, but I want to replace the product. So give me the prompt for the same again. If you, if you don't know how to write those sentences, just ask Jeep to do that for you. Just paste it to nanobanana Pro and your job will be done. But most of the times when you're. When nanobana Pro also struggles because when you have already generated a visual aesthetic, this is one of the most amazing tricks which people still are not using, which is very basic to be honest, but they are not using it. For example, you're still your Nano Banana Pro is also not able to generate the exact same product. Product is way too complex. The, the text or something is not right. Maybe the product overall is right, but maybe there's a line or there's some kind of element that is missing and you are very particular about it. I'll show you one trick. So if you use Photoshop, you can use Photoshop or any other, you know, online available tool. For example, this is one chest band product. For example, my product needs to look exactly like this. And in case Nano Banana Pro chatgpt are not able to do it, I have created the full design. This is just a basic image I'm just showing for the, for the tutorial. So I'll just click on erase tool. I'll just erase that part. Okay? For example, it is able to create this full product, but this part where the label is, it is not able to do that. So I'll just erase this full thing. Okay? And now I'll save this image. I'll save this full image, head over to Nano Banana Pro, upload this image, upload my product, my original product image, Add that to like, add a full prompt. You can ask RGP again for the prompt. And then when you create the image, your product will be exactly the same. It will be exactly the same like 99.9% of the times.

[A]

Awesome. So I want to, because some people are not watching this, I want to explain the process again because this is pure gold and I'll explain afterwards why it works. So let's say you have a product that is complex and it has a lot of things, a lot of components and specific text and so on. And every now and then AI will miss something. It gets everything else right because AI is now good at, like you said, keeping consistency, but it's missing. Let's say in this particular case the text, you can go to Photoshop or PowerPoint for that matter, and delete or Canva and delete the section that is not correct. Then you go back to Gemini and you give it the image with the section that is deleted. Deleted. So there's a white box where that section is supposed to be. And you give AI potentially two or three images, one of the full products, so we know how the whole thing looks. But then a zoomed in section on the section that was not correct in the original image, and the combination of your image with the missing part, together with the full view of the entire product, together with the detailed view of that specific segment, you will get a perfect image. Maybe not the first run, then the second run. The trick for that is when AI tries to create 67 different things accurately, it will not. It will create 53 out of them accurately and the rest are not going to be great. So if you can make it focus on the three things you wanted to get right, which is exactly what ASTA is doing in this process, it will get it right. And so you're going from a create this thing that has 67 different really critical things that I want to reproduce to, oh, this exists. Just fix these three things. It will fix the three things. Love it, love it, love it. It's a great tip.

[B]

Thank you. And I also want to show one example that differentiates the results. Okay, so if you see these images I created purely with the I. Okay. I'm not used any traditional method, but not even just, you know, defining it. I have not refined them. If, see, these are the images, these are all very, very nice Amazon listing images.

[A]

Okay, again, those of you are watching, they're incredible. Like l' Oreal shampoo images on like this dark burgundy kind of background. They look freaking awesome. So now we continue.

[B]

Yeah. So if you see these results, again, why I got these results again from. If I normally use ChatGPT, I would never get these results. I would get genuinely AI slob with so much of text. Because chat GPT's biggest limitation is initially in the starting, if you ask it to generate a particular image, it creates way too like, it adds so much of text, which looks very sloppy. And most of the brands, they always go with minimalistic vibe, very, you know, visually aesthetic, easy to read and, you know, easy for the eyes. So if you see all these images, why I got this kind of result in just a single click because I used very good references. Most of the people still, they believe that, no, we are not going to use any references because that will be just again, copying stuff and, you know, maybe is going to create Something that, that our brand is like. We want to create something from scratch, very beautiful, like everything should be. Nobody has ever seen that on the Internet. You definitely can do that. ChatGPT will not create exact same, you know, visuals that you use as an input, but it will just understand what you have in mind. So you very, very, it's very, very important to, you know, use visuals as references. And for example, if you already have good number of assets for, for your brand. For example, this is one of a very famous product, very famous brand which sells cookies. So I just went to their website, I screenshotted their full website, like how their visual aesthetic is, for example, if this website comes to me, for example, if you already have a brand, if you have already been creating assets for your brand, now if you want to try and experiment with different kind of assets, you can always use them as an input and get completely different, you know, assets and you can genuinely, you know, try and test which one works for your brand. So this can be great for especially, you know, Amazon listing images, Amazon product images. Again, these won't be hundred percent there, but they are still far, far better than if you try to recreate something from scratch. So always try to, you know, input something that really resonates with what you have in mind. And after that, let me also, I

[A]

want to pause you for just one second for those of you who are watching. So those of you who are not watching, I'll explain and those who are watching, I will give you additional information. What we're seeing on the screen right now is a full design process, kind of like a flowchart on the screen. So think about a canvas with multiple things that are leading from one thing to the other. There are multiple tools like this today. Magnifique is one, we've is another. I'm forgetting the third one that a lot of people are using, but they're very common today. And what they enable you to do is they enable you to build a process as a template, if you want. So then you can put in images as references in the beginning and it will do the rest of the steps on its own. So in this particular case it creates the ads, but the assets are visual examples of the style. So everything that Asta explained that you can do in ChatGPT becomes automated. You don't have to do the process time and time again. The process is already built. You, you just have to give it different references in the beginning and you're gonna get completely different outputs at the other side, still following the same process. So this adds focus on this, this adds focus on that. This is the angle on this one, this is the angle on that one. But the reference images are different. So you're gonna get a very different style, different text, different so on depending on the reference images. So instead of doing the process time and time again inside of ChatGPT, you come to a tool like this, like magnifique, like weavy and so on. And I'll put links in the show notes so you can find them easily. You be able to follow to build these processes so you can reuse them instead of manually doing them every single time.

[B]

Yeah, I also want to talk about one of the platforms, it's called Preterm. Like I recently used their platform and it's really, really good. They have launched one AI brain. So what this brain does it like whatever I told you with chatgpt, it kind of does the same thing again with chatgpt you have more control because if you're not liking anything, you can just prompt it out. Okay, this is not something that I'm liking. You just change this, blah blah, blah. So with this, if you like, I have created this brain. So let me show you how this works. So you need to click on this new brain again. Click on this plus icon, add your website URL here and it will start and fetch everything related to your brand, the kind of products you sell. It will, you know, differentiate each and everything. Okay. This website has these different products. This is the visual language. Like it will fetch everything. Okay. And then you create a batch, then you create workflows, basically. Let me just show you the kind of workflows that I created. Yeah, especially if you're an E commerce brand or if you post regularly on your social media, you have to create workflow once and then it will run automatically. So what I did here was this was the product. Like I'm not sure if the audience is aware about the node based platforms, but if you have this built once, you can run them automatically. And this, this will scale your productivity. So you can generate multiple and thousands of assets on every single day basis by just clicking a button. Again, out of those thousand assets, maybe only 200 are useful. But I think you have saved a lot of time so you have to be mindful. Okay. Even if you're creating thousand assets, you need to make sure your inputs are far better. So I've created this workflow. So all you need to do is just enter your product image, add a secondary product image, for example, your packaging or anything that you Want add your website URL, your brand context in case you want. Otherwise AI will automatically do that. Some references, I have uploaded these references that the brand already use and then it. And you can even add, you can take a screenshot of the typography that you want. You can even if you have a project, for example, if I'm trying to create seven listing images for my Amazon store page, you can just, you know, add that simply. You don't have to, you know, be. You don't have to be too specific. But yeah, just randomly say what you're trying to look and it is all connected to Claude. Claude will study your website, analyze each and everything, analyze your references, what are your visual aesthetics and it will write seven different prompts for you that can generate seven different images with just a single click. Again, it would be wrong to say that all of these will be useful in one go, but you are again 60% there. You don't have to create designs from scratch. So once you have this, you can even ask AI. They had, they have a feature called Ask AI and you can, for example, if somebody doesn't know how to create this workflow, right. This is. This seems to be quite complex for beginners especially. So you can just go to ask AI and just simply tell that, okay, this is the workflow that I'm trying to build. I want to input my product assets, I want to build my like input references, I want to input website, whatever inputs that you want to use, just put that and I want the results to be seven different images, seven different prompts and it will automatically build this full workflow for you so you don't have to manually, you know, join nodes and do all of the complex work. So you can easily create such kind of workflows if you have, if you're servicing a lot of brands, for example, for example E Commerce Agency especially you know, handling Amazon listing images for different kinds of brands. If you have these kinds of system in the back end, again with efficient designers who can use these assets and maybe tweak and maybe add 30% of refinement to this, you can create a bunch of assets in a very less amount of time. So you can, in case you don't want to go that ChatGPT route, you can always try and experiment with different platforms out there. So you can create such workflows using Magnific Pleto vv like all these node based platforms, you can create these workflows. So it's quite handy and very effective.

[A]

So awesome.

[B]

Yeah. So again, one more Thing I really want to talk about and this is very important. So again I really want to talk about skills. So I will take you just, I will take 5 minutes of your time because this is something that is very, very important for people to know. So Cloth skills is something that people have not been using. And one trick just to keep it very concise. For example, if you are trying to create Amazon assets and if there are sops about your company or anything, you can always go to Claude and just maybe, or maybe you can go to YouTube and find few of the biggest influencers in your niche who create amazing content regarding for example Amazon listening images. They talk about how to create the best visual assets. Just download all those videos or maybe just transcript all those videos you can go to Node GPT. So Node GPT is something that can help you. It's a free tool. All you have to do is just add your YouTube transcript here and it will just transcribe your full a YouTube video. Like you can create a bunch of transcriptions using this platform. After you are done with that, just add everything to Claude and ask Claude to create a skill file for you. So you will get a Skill MD file that will be an SOP for you. So whatever skills that you usually see, like nowadays everybody's talking about cloth skills. So these are just simple text based files, these are SOPs. So once you train that skill, you have that skill with you. And whenever you need a prompt for maybe seven listing images, this is again a very useful thing. Just ask Lord that use this skill and create give me prompts for 7 listing images and just use Nanobanana Pro or GPT Image 2.0 is far better. Just create 7 images for your brand in no time. So that is one trick. Just wanted to add that because that is very valuable that I personally use on every single day basis.

[A]

Yeah, so it's an awesome trick. It relates to what we talked about before is the ability to take somebody else's genius that explains how to do something in the best way. Use that information and tell Claude to create a skill from it. And now you can reproduce the same process that the best creators that are willing to share how they're doing what they're doing and you can do the same. And Claude will do this for you in seconds.

[B]

Yeah, you can add your skills here and you will be good to go.

[A]

Quick recap of everything we talked about. Digital assets are critical, right? They are the main thing that people engage with online and online could be anywhere, right? This could be YouTube. This could be Facebook, this could be Amazon, this could be your website. This could be any other visual assets that exist out there. They make a very big difference. They need to be consistent, they need to drive people towards engagement and they need to be aligned with your brand so people know it's you and you can look consistent. If you just ask ChatGPT to do it, it will do it, but it's going to look very generic. And the biggest differences we talked about is using visual assets, teaching it your brand, teaching it a style that doesn't have to be yours. You can learn the style from other sources that are working in your industry, explaining what it is that you're trying to do and that's relatively easy to do. You just use English or whatever language you speak in. I voice type almost everything. So it allows me to express myself in a very natural way. I don't have to think about everything I want to say. You can be very sloppy with how you say it, but as long as you're going to explain it proficiently enough, you're going to get significantly better, highly professional results that you'll be proud using and that will actually move the needle. As I mentioned, on a business perspective, Asta, this was absolutely fantastic. If people want to learn from you, follow you, work with you, what are the best ways to do it, where to find you, and so on.

[B]

So as I told you, I create content on LinkedIn and on Insta as well. So I've recently started my Instagram. If people wants to reach out to me, they can always just text me on LinkedIn. That is the best platform. I like to interact with a lot of intellectual people out there. So yeah, LinkedIn is one of the best platforms you can find me on.

[A]

And we'll drop the link for that in the show notes. You can find her easily. Asta, thank you so, so much. This was really, really fantastic. Like I said, like a million small tips in there that people can use to build much, much, much better content. Again, those of you who are listening to this, there's the link to the YouTube channel as well, where you can go and watch what we were talking about. Or if you're in Spotify, you can also do this. Just don't do this while you're driving. And you can actually see the differences between like generic visual content and well produced visual content where the heavy lifting is done once, right? You do the setup once. And yes, it's more than just give it one prompt. There is more work, but you get significantly better results that are actually going to do something versus oh, yeah, I created an asset because I needed to create an asset.

[B]

Yeah. Thank you so much for having me. And I hope whatever tips and tricks I just explained to the audience, I hope they understand what I'm trying to say because most of the things were visuals, so I hope they go to your YouTube and just to watch what I'm trying to say and explain. So, yeah, I hope they found this useful.

[A]

Thank you. Have a good one.

[B]

Thank you. Bye Bye.