The Art of Consistency: A Guide to Using ‘Style References’ and ‘Character References’ in Midjourney

For about almost a year now, we’ve been using Midjourney to sculpt the brand identity and visual language of Medialesson, aiming to convey our external impact effectively. Midjourney has unlocked multiple possibilities for us, turning abstract processes into visual metaphors. However, the inconsistent style and capricious interpretation of prompts posed a challenge. Thus, we endeavored to maintain a relatively clear and consistent prompt. With the release of Version 6, we now have more possibilities. Midjourney introduced Style Reference and Character Reference, which I’m excited to dive into here.

Version 6’s introduction of Style and Character References marks a significant leap forward, allowing users to specify a desired aesthetic or character trait more precisely. By using these new features, we can guide Midjourney to produce images that align more closely with our brand’s aesthetic preferences.

To use a Style or Character Reference, we just add --sref+ [URL] for style and --cref+ [URL] for character to our prompt and include the copied link to the image in Discord, capturing the aesthetic we aim to replicate.

Here is a sample, as we used the prompt before. We always added style in this prompt, by referencing tilt shift, 3d render and isometric diorama.

Now, by right-clicking on the image, we simply copy the image link by selecting “Copy Link” from the context menu to add the style to our next image in Discord.

Now we can use this prompt without referencing the style in this prompt, we also change the whale to a fox and add --sref , also paste the image link of the picture before.

We can also try to change the prompt more without loosing the visual language

Now let’s try to include a character, let’s call him Simon:

After creating the character, we copy the image and add --cref to our prompt. I will keep the style, of the previous images, so --sref+ [URL] stays the same.

Just a quick tip: For the references you can also attach multiple URLs by separating them with a space.

Midjourney now attempts to incorporate a character based on the image without a detailed description from us. It picks up on key features such as the beard, red hair, glasses, and clothing from Simon. This allows us to create visuals that not only match the aesthetic style we’re aiming for but also tell a story with a specific character.

While the Style Reference, in my opinion, has matured enough to be integrated into our brand identity as a valuable tool, the Character Reference does not always achieve the same level of precision. You will always find someone that will look like Simon in the provided images, but something looks always a little bit different. To obtain better results, you can provide a more detailed description of the character, including age, gender, skin color, ethnicity, body shape (you can even add weight), hair, eye color and so forth. Also add more than one Image URL for reference.

However, this doesn’t prevent Midjourney from experimenting with different facial and body shapes or illustration styles of eyes, hands etc.

However the potential for these tools to enhance our brand consistency is immense, and we’re eager to see how further developments will allow us even greater control and creativity in our visual storytelling. For our brand’s recognizability, we’ll definitely be using the Style Reference feature more.

As for creating a mascot for Medialesson: It‘s too early for that. :)