In our creative journey with Midjourney, we have immersed ourselves in the world of character creation to bring to life the perfect representative for our brand «Interesante». Exploring the various features that Midjourney offers, we have created AI characters that not only capture the essence and values of our brand but also have the potential to establish an authentic connection with our audience.
In this article, we share the process behind creating our AI avatar and how we have merged ideas to arrive at our final character. We explore how the diversity of options in Midjourney has allowed us to challenge our beliefs and expand our creative horizons, leading us to make more informed and creative decisions. Throughout this journey, we have discovered how the combination of ideas and images can lead us to find the perfect representation for our brand, and we want to share the learnings, lessons, and discoveries with anyone who wishes to embark on this same journey.
The Process
The Blend option is Midjourney’s standout tool for merging and combining images. True to its name, the Blend tool allows for merging and combining two or more images without the need for a text prompt. It also allows for combining images with a written prompt. We conducted tests using these two functions to understand the model’s capabilities.
Test 1: Blend & Merge tests without a text prompt
This first test was the fusion of the first characters, obtained in our previous article. For this initial test, we let Midjourney only take the characters’ visual references. We didn’t use a written prompt to instruct Midjourney what we wished to obtain.
Trial 1: Using blend tool without a text prompt
Prompt for Midjourney
Trial 2: Using blend tool without a text prompt
Prompt for Midjourney
Trial 3: Using blend tool without a text prompt
Prompt for Midjourney
Test 1: Blend & Merge results without a text prompt
In these initial tests with Midjourney, using the blend tool and only images as references, we could observe the tool’s functionality widely. The results we observed included consistency in the background of each image, a consistent camera angle, and similar lighting. This tool is useful for obtaining more options of the same style of characters. We noticed that with this tool, we began to have a possibility of obtaining character consistency.
An interesting aspect we noticed was that when merging the images without a text prompt, we obtained more unwanted results, presenting bodily deformations, hand and face errors, and inaccuracies in small details. For this reason, from all the images, it was easy to discard the characters that fell within the spectrum of the Uncanny Valley.
Test 2: Blend & Merge tests with a text prompt
The second test involved the fusion of the characters we used previously, but with the difference that we added a text prompt so that Midjourney would consider our expectations for the image outcomes. The prompts were the same for all the tests.
Text prompt for Midjourney
“A full body shot low angle photograph of a 20 year old realistic Latina. She has tanned skin, distinctively curly brunette hair, and warm brown eyes, highlighted by her big eyelashes and prominent eyebrows. She is wearing a minimalist office outfit. She has a high-end smartphone in her hand, a brand-new smartwatch on her wrist. A heart logo with different colors is visible as a pin in her blouse. She is in an indoor office like in the film the devil wears prada. Morning lighting.”
Essential components that the text prompt should have:
- Subject: Latina woman
- Medium of the image: Photograph
- Environment: Indoor office
- Lighting: Morning light
- Color: Without filters
- Mood: Calm
- Composition: Full-body shot
Trial 1: Using blend tool with text prompt
Prompt for Midjourney
Trial 2: Using blend tool with text prompt
Prompt for Midjourney
Trial 3: Using blend tool with text prompt
Prompt for Midjourney
Test 2: Blend & Merge results with a text prompt
When we incorporated text prompts in the second test, the blend tool produced more uniform results among them. In Test 2, the characters resembled each other more than in Test 1. Small details of the accessories are replicated in the three images: the cellphone, the pin on the blouse and the smartwatch. It is clear that using the Blend tool combined with a text prompt, the results provided by Midjourney are more similar to each other, beginning to give us hope of achieving greater consistency in the final character we want to create.
Useful Tips for Creating Your Own Avatar with AI Tools
- Define the desired characteristics of the character clearly before starting with the tests.
- Limit the number of tests to avoid falling into a repetitive cycle.
- Use the tool with text prompts to guide the creation of the character and reduce errors.
- Conduct iterative tests and adjust details after using the tool.
- Experiment with different combinations of images to maintain creativity and variety in the characters.
- Use the Blend tool for exploration and experimentation, not for achieving consistency.
Lessons We Learned During This Process
- Experiment with different combinations: Don’t limit yourself to using the same images over and over.
- Use a clear text prompt: the tool performs better when given a detailed guide.
- Diversity of images: better results are produced when different images are used.
What's Your Reaction?
