With generative AI becoming an arms race towards supremacy, there are things that the technology can not do.
For example, despite learning from massive data sets that include the internet, the whole World Wide Web and beyond, generative AI tools can sometime become too imaginative, that they are bad in creating consistent characters across multiple images they create.
Midjourney finally found a way to properly do that.
In a leap towards enhancing the capabilities of AI-powered image creation, Midjourney introduced an innovative feature aimed at maintaining character consistency across different images.
This development marks a significant step forward in overcoming a prevalent challenge faced by AI image generators, promising to revolutionize the way artists and creators bring their visions to life.
For all this time, generative AI tools lack the ability to recreate their "imaginations," which means that they tend to not be able to make consistent results.
Traditionally, AI image generators struggle to produce consistent characters due to their reliance on diffusion models.
This happens because these AI models generate images pixel-by-pixel based on text inputs, drawing from a vast database of human-created images and associated tags, as learned from similar imagery and text tags in their massive training data set of millions of human created images.
While this method is highly effective in creating diverse and unique imagery, they fall short when tasked with maintaining narrative continuity.
In other words, diffusion models give AI the ability to outstandingly create fascinating whole new pieces content, but they cannot create storyboarding of a film, a novel, a graphic novel or comic book, or some other visual medium where the same character or characters have to move through it and appear in different scenes, settings, with different facial expressions and props.
Here, Midjourney introduces the -cref URL tag, which signifies a monumental shift.
By allowing users to attach a URL of a character image to their text prompts, the tool is finally able to replicate specific character features such as facial characteristics, body type, and clothing across various scenes and settings.
Users can then use --cw to modify reference 'strength' from 100 to 0. Strength 100 is default, and uses the face, hair, and clothes, whereas the strength 0 will only focus on replicating the face.
For more complex references, users even can blend multiple images.
This consistent content generation ability is certainly a game changer.
Since its launch, Midjourney has become a fierce competitor of DALL·E from OpenAI, the tech company that also created the overly-hyped ChatGPT.
And the ability to generate consistent characters across multiple images has been the attention of users, and not just artists and creators.
This time, Midjourney finally delivers that ability, and this has the potential of transforming Midjourney from an ideation tool into a professional asset underscores the evolving landscape of AI in the creative industry.
Midjourney’s update, which can make a significant impact on how characters are depicted in visual narratives, offering a solution to the long-standing issue of consistency, is a testament to the technological progress within AI.
As AI continues to integrate into the creative process, tools like Midjourney are at the forefront, blurring the lines between artificial intelligence and artistic intuition.
The character consistency feature is a glimpse into a future where technology and creativity converge, offering limitless possibilities for storytelling.