As a team of creatives constantly exploring the intersection of AI and advertising, we've been eagerly anticipating Midjourney's latest update. While v6.1 isn't a complete overhaul, it brings subtle yet significant improvements that could reshape our creative process. Let's dive into what's new and how it might impact our work.
Note: all image comparisons are based on the same exact prompts
Enhanced Textures and Reduced Artifacts: The Devil's in the Details
Midjourney has always excelled at creating visually striking images, but v6.1 takes it to a new level. We tested both versions with extreme close-ups of men's faces, and the difference is striking.
V6.0 was already impressive, but the textures sometimes felt forced and artificial. V6.1, however, delivers a level of natural detail that's truly remarkable. The skin textures and eyes, in particular, showcase a realism that could be game-changing for visual campaigns.
Improved Image Coherence: When AI Understands Human Form
One of the most exciting improvements in v6.1 is its enhanced understanding of human anatomy and movement. We put it to the test with some challenging scenarios.
Our synchronised swimmer test revealed a significant improvement in body positioning and overall composition. V6.1 produced noticeably better results - the movements look more accurate and natural, with only minor warping of some faces. In contrast, v6.0 struggled with additional limbs, heads in wrong directions, and people positioned weirdly in the water without much rhyme or reason.
The infamous "hand test" also showed marked improvement. We prompted both models to generate images of two people holding hands. As expected, v6.0 produced too many fingers and merged hands - the usual difficulties we've come to expect. V6.1, while not perfect, delivered much better results. One image did suffer from some finger warping, but overall, the yield was significantly improved. Interestingly, these images also showcased the enhanced colour saturation and contrast of the newer model.
Perhaps most impressive was our action test. We challenged both models with a "sweating man hitting a tennis ball" prompt. V6.0 results screamed "AI fail" with random, exaggerated water splashes, weird facial expressions, and out-of-context tennis balls. V6.1, however, produced images where 3 out of 4 had quite natural motions. The bodies looked correct, and there was a significant improvement in skin texture. While there was still some random splashing, it felt more like a stylistic mist than weird AI mistakes.
To transition into our next topic - personalisation - we ran the same prompt on the 6.1 model with the added "--p" parameter. The results were quite visually striking, feeling more stylised and cinematic compared to the previous images. This seamless shift from improved coherence to personalisation showcases the versatility and potential of Midjourney's latest update.
New Personalisation Model: AI That Speaks Your Visual Language
The introduction of the "--p" parameter is perhaps the most intriguing feature of v6.1. It's not just an incremental improvement; it's a new way of interacting with AI. Although Midjourney hasn't fully disclosed how their personalisation model works, they've revealed that it mainly runs on the images you rate and like on Midjourney to gauge your style.
We tested this feature with two prompts. First, "2 people posing in the middle of the desert." Without personalization, we got basic images reminiscent of 2015 Instagram Coachella aesthetic. With the addition of "--p", the results were much more unique, curated, and artistic. While not perfect, these images cater more to our personal style and showcase the power of personalisation.
Our second test, "cinematic shot of two people having coffee," really demonstrated the feature's potential. Without "--p", we got cookie-cutter images with darker lighting. With "--p", the results looked straight out of an art film. These could easily serve as cinematography reference points or even inspire a script treatment.
The best part? This heightened personalisation saves time by reducing the need to pile on style descriptors in our prompts. It's like having an AI that truly understands our aesthetic preferences.
Text Accuracy: When Every Word Counts
For those of us in advertising, accurate text rendering is crucial. V6.1 shows promising improvements in this area.
We started with a simple prompt: "a billboard that says welcome to hell". V6.0 had a decent amount of mistakes, with only 1 out of 4 looking quite good. V6.1, however, achieved 3 out of 4 accuracy across generations with far fewer mistakes.
To really push the limits, we tested a more challenging phrase: "Nor here nor there but everywhere". The repetition of similar letters and words proved challenging, but v6.1 managed accurate renditions in 2 out of 4 generations. V6.0, on the other hand, couldn't quite make the cut. While not perfect, this improvement could be a game-changer for brands using AI, potentially reducing the amount of post-generation work.
The Future of AI in Advertising
As we continue to explore the capabilities of Midjourney v6.1, it's clear that AI is becoming an increasingly powerful tool in our creative arsenal. However, it's crucial to remember that AI is just that – a tool. The real magic still lies in the creative minds wielding it.
These improvements in Midjourney v6.1 don't just represent technical advancements; they represent new possibilities for creative expression. As AI becomes more nuanced and personalised, it allows us to push the boundaries of what's possible in advertising and design.
But as we embrace these new capabilities, we must also consider the ethical implications. How do we ensure transparency in AI-generated content? How do we maintain authenticity in an era of hyper-realistic AI visuals?
These are questions we'll continue to grapple with as we integrate AI more deeply into our creative processes. For now, we're excited to explore the new frontiers that Midjourney v6.1 opens up. It's not about replacing human creativity, but augmenting it – allowing us to bring our wildest ideas to life faster and more vividly than ever before.
Commenti