AI tools are gradually improving. They now allow not only experimentation with text or visual content but also the creation of full-fledged AI bloggers. In this article, we discuss how to create your own virtual character that can help to promote offers.
The Emergence of Virtual Influencers
It was in the early 2000s when computer-generated fictional characters began to emerge in entertainment industry: older generations remember the debut and rise of the popular band Gorillaz. The anthropomorphic cartoon singer Miku Hatsune, with her synthesized voice, appeared in 2007.
The Instagram account of the virtual 3D influencer, model, and singer Lil Miquela appeared in 2016, and for a long time, hundreds of thousands of her followers didn’t even suspect that they were following the life of a fictional character.
In recent years, a unique sector of the entertainment industry has emerged around these and other virtual characters making it possible for developers of non-existent “celebrities” to earn money. This is a CGI industry, which stands for “computer-generated imagery.” With the development of neural network technologies, the process of creating and promoting CGI influencers has been refined. As a result, overall costs have decreased. Now, a company or marketer with a relatively small budget can create their own virtual blogger. Regular users can create their own models for free using special services.
Famous AI Influencers Worldwide
The economy of the CGI industry is constantly growing. For instance, in 2019 businesses spent $8 billion on maintaining and creating CGI characters, in 2020 it was already $10 billion. By 2022 companies invested over $22 billion. What is interesting is that some AI bloggers “create” more engaging content than real people. The average engagement rate for a CGI model is 0.7%, while for Lil Miquela, it is 2.7%.
In 2009 Lou do Magalu appeared as the mascot of the Brazilian online store Magazine Luiza. In the early days, Magalu helped customers choose gifts from the catalog, but soon became so popular that she became the official face of the brand. Nowadays, the developers of Magalu are approached by globally recognized companies for collaborations, including Adidas, Red Bull, and Samsung.
Competition in the CGI character market is only intensifying. In November of last year, a new star emerged — pink-haired Aitana Lopez, created by the Spanish modeling agency The Clueless. Within just a few months, her Instagram account gained over 121,000 followers, with thousands of comments appearing under her posts every day.
Around the same time, AI influencer Emily Pellegrini quickly gathered 149,000 followers on Instagram within a few weeks. Emily’s creator chose to remain anonymous but told the Daily Mail that he designed her sexy appearance with the help of the ChatGPT chatbot. After the project launched, Emily began receiving dating invitations in her DMs, including from well-known and wealthy individuals on the Forbes list —millionaires from the UAE, football players, MMA fighters, and tennis players.
In March 2024, the model Samantha Everly appeared on the cover of Mexican Playboy, her image entirely created by neural networks for generating images. She was created by the XChatter team, and landed on the cover of the famous magazine after gaining popularity online. She has accounts on several social media platforms, where she “communicates” with followers on a variety of topics. The total number of her followers exceeds 119,000 users.
How much do AI influencers earn?
The total audience of the AI influencer Lou Magalu across all popular social media platforms associated with the activities of Magazine Luiza reaches 30 million users. The cost of a sponsored post featuring Lou is estimated at a record $20,000 for the industry.
According to the British marketplace OnBuy, one of the wealthiest CGI influencers, Lil Miquela, earns more than $11.7 million for her creators, with one post on her Instagram costing $10,000. Miquela’s primary source of income comes from brand collaborations with Balenciaga, Prada, and Fenty Beauty.
Aitana Lopez earns money promoting dietary supplements and sports nutrition from leading brands. Her photos in lingerie also regularly appear on Fanvue, a platform similar to OnlyFans. According to Ruben Cruz, founder of the agency The Clueless, the model’s average monthly income is $3,000, and in the best months, this amount can reach $10,000.
The AI persona Emily Pellegrini generated more than $10,000 for its creator from the sale of explicit photos on Fanvue after the project was launched.
How you can make your own AI influencer
The Virtual Humans database has several thousand contacts of AI influencers, each with more than 1,000 followers. Experts claim that the majority of the characters listed in the database either abandoned their accounts or manage them irregularly. A typical post from a virtual blogger usually gathers no more than 10 comments, and the engagement rate doesn’t exceed 1%. This means that many digital bloggers are no more popular than an average user who collects no more than 200-300 followers by the age of 30 from friends, acquaintances, and family.
Creating and promoting a truly successful AI influencer requires investment. However, by using some accessible services, anyone can create their own AI blogger. Let’s look at several popular services.
RenderNet
To register on the website, you can use an active Google account or an email address.
After registration, you get access to the dashboard with an intuitive interface. When you start, you get 10 free credits which is enough to test the main features and see if it’s worth working with RenderNet AI.
To create an AI character or a person, you should click on the “Studio” button in the left menu, which opens the toolbar panel.
Select a model type from the “Selected Styles” set. There are various customization options to personalize your character:
- Quality: you can choose the quality of the future image — either 512 px or 768 px.
- Aspect Ratio: it is possible to choose the aspect ratio: 1:1, 2:3, or 3:2.
- Models: this menu expands the style selection options. More than 40 variants are available to choose from.
- LoRA: this section includes additional components that further personalize the creation. For example, adding the Food Photo LoRA incorporates food-related elements into the composition. The Better Hands LoRA allows experimentation with the model’s limbs. All LoRAs in RenderNet can complement each other, increasing the variety of the final result.
- ControlNet: this tool allows you to upload an image, and the neural network will replicate a specific pose or body position based on it. Additional settings for this parameter let you experiment with visual effects. You can generate an image in photographic negative, adjust similarity to the original, scale, and more.
- Settings: in the general settings, you can adjust technical parameters such as the starting point for generating the prompt, the sampling method used during rendering, the number of steps for the sampling process.
After selecting the appropriate settings, you should write a prompt and, if necessary, enter “negative” prompts in the Negative Prompt field to exclude certain elements from appearing. The image generation takes 10–20 seconds, and RenderNet provides four photo options as output:
After creating and saving the image, go to the FaceLock tool, which allows you to generate different images of people while preserving the given face. FaceLock works most effectively with close-up faces, free from shadows and glasses. To activate FaceLock, click the toggle and switch the function to ON, then simply upload the desired image into the workspace.
You can access the FaceLock menu immediately after saving the photo.
After updating the prompt in FaceLock, you will receive a batch of new images with the specific type of face you’ve selected.
The free version gives each user 10 credits — one credit for each generated image. Credits can be purchased: the larger the number of credits purchased, the higher the discount. The service can also be used for free by registering new accounts — when a new user logs in, they are awarded the same 10 credits, which are reissued 24 hours after each use.
Fooocus Stable Diffusion XL
This is a neural network for generating images from text descriptions, based on the popular Stable Diffusion XL. Its developer is the creator of the well-known ControlNet extension, which helps to generate images based on an existing picture, sketch, or pose.
The Fooocus Stable Diffusion XL service features a GPT engine that helps expand the context of input prompts to generate more detailed images. Additionally, it supports the FaceSwap function — perfect for creating non-existent photo models, bloggers, and personal AI influencers.
To install Fooocus Stable Diffusion XL on your PC, you will need:
- A computer or laptop with a graphics card that has at least 8GB of video memory.
- 25GB of free disk space for one launch mode, and 40GB for all three modes available in the service.
Installing Fooocus Stable Diffusion XL on Windows is simple:
- Download the archive from the provided link, extract it, and run the run.bat file.
- A console will open on your computer displaying the installation process, and it will automatically download 6.5 GB of required files.
After the installation is complete, the neural network interface will open in your browser, allowing you to work with the service. Upon launch, you can choose from one of the modes with various models, appropriate settings and styles:
- General: a universal mode that is launched using the run.bat file.
- Realistic: a mode for creating photorealistic portraits, launched with the run_realistic.bat file.
- Anime: a mode suitable for creating art and anime, launched with the run_anime.bat file.
When the program is launched in realistic mode, the system sets optimal settings for creating realistic photo portraits. You may need to adjust the image size, and you can also enable the Quality option if your PC’s power allows for high-quality image creation.
To access the sidebar with settings, check the box in the Advanced menu.
The optimal image size for generation is 1024×1024 (1:1). Other settings can generally be left unchanged. In the Style tab, you can choose from various styles instead of just realistic or anime. There are dozens of options available, including exotic ones such as Futuristic Retro Futurism, Adorable 3D Character, Steampunk 2, and others.
The number of generated images can be adjusted using the Image Number slider. To generate an image of a female blogger’s face, create a prompt. Since Fooocus automatically expands the context of the request with its built-in GPT, you can enter simple prompts like: “young beautiful girl with blonde hair and pigtails.”
The generation process takes no more than two minutes, after which the service will provide a selection of images to choose from.
To save any image, you can either left-click on it or click the icon on the image itself. As a reference instead of a text prompt, you can upload your own image by clicking the Input Image button.
After working with the image, you can improve its resolution, as 1024×1024 is the minimum size and may not be suitable for social media. To enhance the resolution, use the Upscale or Variation tool in Fooocus Stable Diffusion XL. To do this:
- Go to the Upscale or Variation tab and upload the previously generated image.
- In the settings, select the Upscale (2x) mode and click Generate.
Upscaling not only increases the image’s details but also boosts its resolution.
Conclusion
Creating your own AI character and promoting offers in active social media channels using unique visuals is not that difficult. The services we discussed allow you to do this for free, but developing and creating content will still require a certain amount of time and effort.