Despite continuous advancements in AI Tools that have shown promising results in various areas, affiliate marketers have often been skeptical about using them to generate ad creatives.
In this article, we will analyze 9 AI tools and attempt to answer the question: "Is it worth using AI Tools, or is AI incapable of producing genuinely high-quality creatives?". If you want to find out more information on AI Tools and how to work with them, visit our Telegram channel.
Midjourney
The Midjourney AI tool for image generation is well-known and has demonstrated impressive results. We decided to assess the quality and appeal of the creatives it can generate and whether they are suitable for marketing purposes.
At present, Midjourney is accessible to anyone, regardless of the power of their PC or laptop. Since not everyone may be familiar with how to use it on Discord, we will provide a brief guide on how to get started.
Registration and Getting Started
Go to Discord and join the official server. After joining, you can choose between two options: generate in the general chat or create your own chat room. In the former case, your creations will be visible to all chat participants, while in the latter, they will be visible only to you.
We will generate content in the general chat. Click on the room for beginners, whose names start with "newbies":
After confirming your account in the selected room, you will see a prompt input field. You can now start working on your creative project:
If you happen to lose a generated image, simply access it in the mentions section:
To initiate content generation, you need to familiarize yourself with basic commands, each of which begins with "/." Here are the most important commands and what they mean:
- /imagine: This is the primary command that directly generates an image;
- /setting: This command opens the generation settings window;
- /info: It displays information about remaining generations or purchased hours;
- /blend: This command allows you to blend two images into one.
Additional buttons are available beneath the generated image, allowing you to modify or enhance the output:
Here are their functions:
- U: Enhance image quality, where the numbers represent which of the four options to enhance.
- V: Variations using one of the four images.
- The blue button with arrows on the right signifies regenerating based on the same prompt.
Please note that each time you press any of these buttons, one generation is consumed.
Generation and Writing Prompts for Nutra
Now that we are ready to get to work and understand all the basic functions, let's begin experimenting. Based on what the AI tool is capable of, we decided to attempt generating creatives for Nutra and gambling niches. Since the AI refuses to create images of naked women, its suitability for adult and dating content is limited.
We will try to generate creatives for Nutra in currently popular ad styles: parasites, enhancers, and weight loss. When writing prompts, specify not only what should be in the image but also the style, colors, spatial arrangement, quality, and so on. Here is the first result where we asked the AI tool to generate a girl with worms:
It is undeniably creepy, which is precisely what Nutra demands. Such images trigger the urge to purchase the product, just to avoid encountering such monsters in reality. By the way, the use of the word "parasites" is prohibited by the rules, so to generate it, use the word "worms" instead.
Direct references to intimate body parts are not allowed. Therefore, we decided to draw an analogy with a banana, and here's what this AI tool had to offer:
Clearly, the AI tool does not grasp what is being asked of it, resulting in a less-than-ideal outcome.
We also attempted to create a creative for green smoothies, which is also popular. We wrote the prompt "beautiful girl drinking a magic green smoothie in the middle of the apartment in a realistic style detailed image":
Additionally, there is an AI tool for helping craft proper prompts called Dall-E 2 Prompt Optimizer. We decided to test it as well since it supposedly enhances generation accuracy and helps achieve better results.
On the main page, we are asked to input our prompt and then select desired parameters:
At the top, you can see the original text, and at the bottom, the generated prompt. We used a slightly different text as the prompt, but it was still about green smoothies. We immediately input the refined prompt into MidJourney:
The service is helpful, but it only assists at the beginning. Essentially, it merely adds selected words from the list. The second result was also interesting but did not meet the creative requirements, as it did not show the result after consuming the smoothie.
We also decided to generate an image suitable for the gambling verical. In one request, we asked to combine several popular casino games. Here's what we got:
MidJourney's results turned out to be cool and high-quality. It can easily be envisioned as the closing image in a classic gambling video creative for Facebook.
In conclusion, MidJourney is an excellent choice for creative content creation, especially in the Nutra niche. Furthermore, if you refine your prompt-writing skills, the results can be impressive.
Newcomers are granted 25 initial generations, which are enough to explore all the functions and determine whether it's worth investing in the paid version. Afterward, you can purchase the basic or standard version. The first option costs $10 per month and provides 200 GPU minutes, while the second is priced at $30 per month for 15 GPU hours. In short, it's more cost-effective to opt for the standard version upfront.
ERNIE-ViLG 2.0
Recently, Chinese developers introduced their Text-to-Image AI tool. According to their claims, it combines the capabilities of MidJourney and Dall-E 2, and it excels in generating anime characters, although we have little interest in that aspect. Let's try using it to generate creatives for Nutra and compare it with MidJourney.
The AI tool is available for free online. Each request takes approximately one and a half minutes to generate, resulting in four variations.
For the first test, we used the same prompt as in the first green smoothie generation. Let's see what the Chinese AI tool came up with:
The first image suggests that the smoothie is, to put it mildly, not tasty. In the second image, the straw became part of the head, but the last two are not bad and could serve as a background for a creative.
Now, let's attempt to generate an image for advertising a product against parasites, similar to what we did earlier. The prompt is identical, but this time we used the word "parasites." Here's the result:
Unfortunately, the AI tool did not come close to understanding what was expected of it. We tried generating it again in the hope of a better outcome and removed the word "scared" from the prompt. Then the AI tool got completely confused:
The advantage of this AI tool is that it is free to use, and there's no need for any preliminary setup. Moreover, it occasionally produces decent results. Users primarily employ it for creating anime-style content, and if that's precisely the style you require, it works excellently.
A small note: the AI tool does not generate some political queries and Winnie the Pooh.
Pornpen.ai
We previously discussed Pornpen.AI in September 2022, when it was just gaining traction and lacked significant popularity. At that time, the results of its generations were more amusingly eerie than suitable for creating creatives in the adult and dating niches.
Time has passed, and Pornpen.AI has taken the internet by storm with its enhanced version that astonishes users with its realism. Some people even thought and argued that the pictures were not AI-generated at all but from real women, while others tried to convince them otherwise.
Following the latest upgrade, the AI tool was trained to create men and various parts of their bodies interacting with women. Previously, the AI tool exclusively generated women only:
Moreover, the number of prompt parameters for generation has increased four times compared to before. Now, you can create absolutely any variation that comes to mind. There are more options for backgrounds, clothing, hairstyles, body features, ethnicities, lighting, and actions. While previously, a woman could only stand, now she or they can perform actions:
Results can be tailored to your specific needs, or you can choose from the constantly emerging generations by other users. Most of the impressive results we've achieved cannot be showcased, but here are a couple of pleasing visuals:
You can use the AI tool online for free without restrictions, but you can also opt for a Pro subscription, which unlocks additional capabilities. This includes fixing and enhancing individual parts of the image, adding other elements, improving image quality, and more.
This subscription costs £14.5 per month.
Photoroom
The AI tool-based program Photoroom doesn't generate images but excels in refining them with high quality. It offers a multitude of useful functions for creative content creation. For example, it can remove, replace, or blur backgrounds, as well as act as a magical eraser for unwanted objects within an image.
Let's test Photoroom's capabilities using an example from one of the creatives we created in Pornpen.ai. First, we upload the photo to the editor and click the "Background Remover" button:
The result is outstanding. Take a look at how well the AI tool handled complex objects like hair.
Additionally, the watermark from the previous AI tool disappeared along with the background, eliminating the need for additional steps. Now, let's try different indoor backgrounds and see how realistic they appear:
The result already looks pleasing, and such an image would be suitable for social media platforms like Instagram*. By the way, you can immediately select the specific social media platform for which the post is created and add shadows and reflections for added realism:
There is also an option to upload your own background from the library. We decided to try creating a beach photo. To make it look realistic, we took a real photo of a girl on the beach and then removed the girl, testing how the eraser function works in the process:
This result was achieved in literally one click, and the processing took less than 5 seconds. This is excellent, considering that we will overlay our generated girl on top. After some adjustments to the photo, including slight changes in body orientation, we get this result:
If used for creatives or avatars, the result is simply excellent. Of course, upon closer inspection, the absence of a sunlit shadow on the sand, which Photoroom cannot add, and color discrepancies become noticeable.
Speaking of how well Photoroom serves people selling products online, the answer is unequivocal: the service is ideal. The application offers numerous marketing tools and backgrounds for beauty products such as creams, perfumes, and more:
Recently, Photoroom introduced a feature called Magic Studio for generating promotional images. The essence is as follows: the AI tool replaces the object's surroundings with a scene generated based on a textual prompt. You can see how it works below:
All the prior features can be used for free and without limitations, with the option to purchase the Pro version available. You can use Photoroom online here. Additionally, there is an app available for iPhone and Android. The Pro version costs $9.99 per month, while an annual subscription is priced at $69.99:
The Pro version includes the removal of the application's watermark, saving in high resolution, image optimization for specific marketplaces, and the ability to work on multiple devices.
In summary, Photoroom is an excellent application that will be useful for both affiliates and advertisers.
eDiff-I by NVIDIA
Recently, NVIDIA introduced its latest development in the field of artificial intelligence - eDiff-I. Its peculiarity lies in the fact that to create an image, you need to make a very rough sketch and break it down by colors. This way, the AI tool should understand the shape, color, and compositional placement of objects. As a result, the user obtains a more accurate image. Here are a few examples of how it works:
At the moment, eDiff-I is not publicly available and is being tested within the working team. The release date for public access has not been announced yet, but we will be eagerly awaiting it!
Runway
Runway is an AI tool developed by Stable Diffusion. Its main function is that it can change a selected element in a photo or video. All you need to do is highlight the area and write a prompt for replacement:
For videos, there are functions such as motion tracking and smart object removal. However, today, we are only interested in working with images.
On the homepage of the website, you can see all the AI tool's functionality, but let's start by testing "Erase and Replace":
We'll begin with a simple task: adding a turtle to the sand next to a girl. We select the area where the turtle should be, write a prompt, and click the "Generate" button:
Runway immediately offers 4 options for the object, and the user only needs to choose the most harmonious one:
The turtles look quite good, especially in the second picture. Now let's try drawing clothes on the girls we generated earlier. We selected the area and wrote a prompt, specifying that we need a white t-shirt:
It turns out that Runway does not accept or process images of this nature as it is considered inappropriate content. Therefore, it's not possible to process adult publications here.
Let's try adding a large eggplant to a pair of men's pants to create a creative for the adult niche. We took an image from a search engine. Runway produced the following result in just a few seconds:
The result turned out like this, it's not even an eggplant but a zucchini. It looks amusing, and perhaps such an image can be used for creatives. And if you add a call to action like, "Your zucchini will grow like wildfire," what user can resist?
Runway is a paid program, but newcomers are given 25 generations for testing. After that, you'll need to purchase a Standard subscription for $15 per month, which includes 625 generations, or a Pro subscription for $35, providing 2,250 generations.
Pulse
Pulse is a recently released AI tool that transforms pixelated face photos into high-quality images. It's important to note that Pulse doesn't recreate the actual face but generates a fictional one that fits the context.
In the early days after its release, some were concerned that Pulse could be used for deblurring intentionally blurred faces, but developers say that the AI cannot be used for such purposes. Here's an example of how Pulse works:
Pulse is a free program, but it needs to be installed on your PC to work. You can download it here.
Typefully
Typefully is a modest AI tool that allows you to create beautiful designs for social media. It's very simple to use: you need to upload an image and then adjust its appearance based on several parameters:
Here, you can adjust the color, the size of the outer and inner frames, corner rounding, add shadows, gradient background, and positioning. Typefully is mostly tailored to the Twitter format, but affiliates can use it to create and design landing pages or websites, for example:
Typefully is completely free, and images are saved in high quality.
ControlNet
ControlNet is an impressive extension for Stable Diffusion that allows you to connect to its individual blocks and edit them. This enables us to create images of an entirely different level. ControlNet involves working not with a textual prompt but with a reference image, in other words, Image-To-Image.
The extension can identify object boundaries and convert images into sketches, create depth and normal maps, determine the poses of characters, and generate images based on a sketch:
Thus, all these functions allow you to create fully custom images based on a reference and a sketch, as well as combine two images together:
In theory, this extension, together with Stable Diffusion, can be used to create completely unique, high-quality creatives. You don't need people, equipment, post-processing, or anything else for this.
The only downside is that setting up ControlNet requires a fairly powerful PC with a graphics card of at least 4 GB, but the recommended value is 8 GB. Most functions will work properly with these specifications. The weight of the extension itself, in addition to Stable Diffusion, is 40 GB. You can download the extension here.
Conclusion
Some of the programs listed have been around for a while, while others have just been released, but each of them can be useful for making ad creatives. AI Tools are primarily relevant for the "Nutra" niche, as creatives for these offers tend to have a somewhat imaginative character. They can be useful for adult offers, but their utility for gambling offers is limited.
We hope that some of these AI Tools will help affiliates find fresh ideas and make their work at least a little easier. Good luck!
