With the rise of Artificial Intelligence (AI) tools, at Interesante we decided to create our own character completely made with a text to image model. Our vision stems from the fact that an Avatar will allow us to have a distinctive and recognizable visual identity. Having an avatar that represents Interesante on digital channels makes it easier to personify the values and the personality of the brand. Aspects that often go unnoticed by the public, but are crucial for companies to understand it clearly.
The mission of the avatar is to have a human brand, a brand that is accessible and friendly, giving an image of coherence in different platforms and increasing its recognition in the audience. Clear examples of characters that represent a brand, are Ronald McDonald’s clown, a distinctive emblem of the restaurant chain “McDonald ‘s” or Santa Claus, who represents Coca-Cola since 1931.
These characters work so well that just seeing them lets people know which brand they belong to, without needing any additional visual input. Clearly, the characters (avatars) have always been present and are a powerful tool for branding. That is why, in Interesante, we decide to go down into the rabbit hole and begin this adventure towards the unknown of creating a character with Artificial Intelligence tools.
Midjourney, one of the most well-known Artificial Intelligence models in the design and illustration field, has entirely changed the game for image creation. Its text-to-image model simplifies creating any kind of image using just a text prompt. This tool is capable of generating images in a kind of composition such as photographs, images and illustrations, so creating a character must be within its capabilities. We aim for Midjourney to produce a photograph of a person who is as authentic and realistic as possible, and can be replicated across various situations.
The process
For the creation of our avatar, we decided to generate a brainstorming session, where we defined the vision we wanted to create, the purpose this avatar would serve, why we needed it, and how it would support our brand. In the creative process, it was essential to free ourselves from all prejudice and gather as many ideas as possible to help define the profile of our avatar. Based on the values, mission, and vision of our brand, Interesante, it was simple to ground these ideas, have solid foundations and start feeding the tool with all our ideas.
Along with Midjourney, we decided to use another Artificial Intelligence model known as ChatGPT to assist us in generating ideas and guide us on how to start creating our Avatar. ChatGPT is a large language model that uses artificial intelligence to generate coherent and contextually relevant responses to what is indicated. This model became quite well-known in 2023, as it aids in various tasks such as creating content, being a virtual assistant, and aiding in the research process. It also serves as an analysis tool and a source of innovative solutions, so we decided it was essential to include it in our project.
We began our character building using a prompt known as Prompt Creator. The goal of the Prompt Creator was to introduce it into ChatGPT and for it to return a complete guide of everything we needed to consider for generating our first images with Midjourney.
Prompt Creator
I want you to become my Prompt engineer. Your goal is to help me craft the best possible prompt for my needs.
The prompt will be used by you, ChatGPT. You will follow the following process:
1. Your first response will be to ask me what the prompt should be about. I will provide my answer, but we will need to improve it through continual iterations by going through the next steps.
2. Based on my input, you will generate 2 sections, a) Revised prompt (provide your rewritten prompt, it should be clear, concise, and easily understood by you), b) Questions (ask any relevant questions pertaining to what additional information is needed from me to improve the prompt).
3. We will continue this iterative process with me providing additional information to you and you updating the prompt in the Revised prompt section until I say we are done.
Please ensure that each time a prompt is created, it should not include quotation marks.
Following this initial input, ChatGPT presented us with questions that needed to be addressed to develop the AI persona fully. The questions provided by ChatGPT were as follows:
- Do you have any preferences for the appearance or visual representation of the AI influencer?
- What specific traits or personality should the AI influencer character possess?
- Should the Avatar have a backstory or personal narrative?
Once ChatGPT gave us some guidelines, we started brainstorming what we wanted and needed this character to portray. We aimed to craft a character with a comprehensive personality relatable to a broad audience.
Our Avatar:
- Is female.
- Is about 25-30 years old.
- Has Latin characteristics: tanned skin, brown hair, curly hair, brown eyes, prominent eyelashes, thick eyebrows, and a normal build.
- Reflects the values, mission, and vision we already have in «Interesante».
With all the details of our AI character established, we set out on our adventure with Midjourney. Conducting a series of tests taught us the ins and outs of the tool, how to secure our desired outcomes, and the way to achieve consistent results. Below, we share our discoveries.
Generating images with Midjourney
To generate images with Midjourney, we recommend focusing 100% on the physical characteristics and visual elements you want to achieve. Directions like “good person,” “kind,” “curious” won’t get you anywhere since there is no visual representation of those words.
After some tests, we learned that it is necessary to include a «negative prompt» for each image generation iteration. This prompt allows Midjourney to make fewer mistakes and deliver a more accurate result. The model is still under construction and learning continuously, so if we do not apply this indication, there is a higher probability of images with mutations, disfigurements, duplicated parts, etc. This could lead us to the Uncanny Valley, something we want to avoid at all costs in our characters, as it increases the possibility of distrust among users, and they would immediately associate it with the brand.
Negative Prompt is used in every iteration
–no 3d, render, CGI, dool, painting, fake, cartoon, 3d modeling, worst quality, deformed, malformed, bad teeth, bad hands, bad fingers, bad eyes, long body, blurry, duplicated, cloned, duplicated body, duplicated body parts, disfigured, extra limbs, fused fingers, extra fingers, twisted, distorted, malformed hands, mutated hands, mutated fingers, conjoined, missing limbs, bad anatomy, bad proportions, logo, watermark, text, copyright, signature, lowres, mutated, mutilated, artifacts, gross, ugly
Test 1 created with Midjourney
ChatGPT’s Question:
Do you have any preferences for the appearance or visual representation of the AI influencer?
Prompt for Midjourney
«Photograph of a Latina woman. Tanned skin, curly brown hair, and brown eyes, highlighted by glasses. Her facial features include large eyelashes and prominent eyebrows.»
Midjourney results
Test 2 created with Midjourney
ChatGPT’s Question:
What specific traits or personality should the AI influencer character possess?
Prompt for Midjourney
«A 26-year-old Latina, tech expert and passionate about environmental activism, confidently poses in a tech and eco-friendly setting, representing the link between the Latin and American worlds.»
Midjourney results
Test 3 created with Midjourney
ChatGPT’s Question:
«Should the AI influencer have a backstory or personal narrative that aligns with their mission to promote technology for women?»
Prompt for Midjourney
«Follow the journey of a woman in technology, from her beginnings in a co-working cafe to becoming an expert in social media and web development, confronting gender disparities and creating inclusive spaces.»
Midjourney results
Profile Refinement
In this final test, we combined the best parts of previous prompts into one. This final prompt used the words that worked best. We also received help from ChatGPT for a different version of the prompt.
Prompt for Midjourney
«A Latina woman passionate about technology and journalism confidently poses in modern attire, with a multicolored heart pin and the word «Interesante» on her blouse, in a coworking space with tech magazines.»
Midjourney Results
Useful tips when creating an AI persona
If you decide to join us on this journey of creating a character for your brand, we recommend following the structure of our creative process. Gather with your team for a brainstorming session, research your target market, reuse the information you already have about your brand, and enjoy exploring this artificial intelligence model.
Here, we give you a series of elements that will serve as shortcuts that you should focus when drafting the prompt you want to give to Midjourney:
- Determine the subject: man, woman, child, animal, alien
- Medium of the image: photography, painting, illustration, drawing
- Environment: indoor, outdoor
- Lighting: soft, warm, studio
- Color: black and white, sepia, filters
- Mood: calm, energetic
- Composition: close-up, full body, aerial shot
Additionally, remember to add specific parameters to style your image.
- Aspect ratio (9:16)
- Style: raw
- Version v 6.0
Tip: These specific parameters are added at the end of your prompt by writing a double slash followed by the parameter abbreviation. To understand in detail all the available parameters and their use, we recommend reading the official Midjourney’s documentation, which you can find at the following link.
Lessons that we had during this process
- Language models are powerful tools that can elevate your brand to new heights at a relatively low cost.
- There needs to be cooperation between the outcomes provided by ChatGPT and the final instructions you wish to give to Midjourney.
- To ensure accurate results in Midjourney, clear parameters must be provided.
- Midjourney filters out unnecessary instructions, making trial and error essential to become familiar with the model.
- Fewer instructions yield better results.
- Broad instructions produce broad results.
- Overly long instructions lead to unexpected outcomes.
- Including «negative» instructions refines the final image, ensuring it aligns more accurately with the desired result.
