For a even though, I have felt like the AI image generation area is turning into as well stagnant. There weren’t any important developments soon after DALL-E 3, and newer designs have left me disappointed or feeling a bit neutral.
A single point stored me going for months: the release of Midjourney V6.
And when it lastly came, it came as a sudden Christmas existing. No a single actually anticipated V6 ahead of 2024, so this early sneak peek of its base model — even with out some of its functionalities — was a welcome shock.
It really is been a couple of days and I am now prepared to give my preliminary ideas on this new model. In this write-up, I will outline each improvement and adjust that came with V6.
What is New With V6?
Midjourney V6 is the AI picture generator’s most current model, expanding on the abilities of preceding iterations and shifting some of its core functionalities. Like with V5, this model is not the last model of V6, rather just the base model which will be steadily fine-tuned in the subsequent months.
Nevertheless, it is nonetheless the most capable Midjourney model to date, with additions this kind of as:
- Enhanced output creativity.
- Greater prompt comprehension.
- Robust upscalers.
- Text generation.
V6 can now get extended and complicated prompts, and produce correct photographs. It comes at a value although, with loyal Midjourney consumers currently being asked to relearn how to prompt given that the model is incredibly delicate. I’m getting the very same concern myself but, at the finish of the day, it is a tiny value to pay out.
How Are They Enhancing The Model?
Apart from expanding its education set, Midjourney is also gathering consumer view to increase its model by way of A/B Testing. This is voluntary but consumers can get totally free hrs if they participate, driving much more consumers to fee photographs on their site.
Midjourney V6 Output Good quality
Now that you have been launched to the enhancements Midjourney produced on its model, it is time to see it in action. Right here are some output examples from V6, grouped neatly by picture group:
Realism (Portraits)
Prompt A: a younger female attending a music festival, backlighting, portrait
Prompt B: a physics professor educating his class, academia, shut-up
If you have been following along with my Midjourney reviews, you know that I’ve extended been annoyed about its tendency to produce waxy faces and overemphasize particular characteristics. With V6, that is turn out to be a point of the previous.
The two of these photographs seem extremely actual. Even if you seem closely, there are no clear indicators that these are created with an AI at all. These examples actually communicate to how far Midjourney has come from V5.two in just a couple of months.
Individual Score: five out of five
Realism (Landscape)
Prompt A: solitary stone cabin in a huge alpine meadow, wildflowers in bloom, snow-capped peaks in the distance
Prompt B: misty autumn forest path, fallen leaves carpeting the ground, sunlight filtering by way of the trees at daybreak
This is one more excellent score for me. I experimented with to journey Midjourney up by utilizing conversational language but its enhanced coherence permitted it to fulfill each word in my prompt. As for the good quality, they are brilliant and vivid with out going overboard, display no signal of rendering problems, the shadows make sense, and the depth of discipline is steady.
Individual Score: five out of five
3D Renders
Prompt A: a minimap diorama of a quiet chic library adorned with indoor plants
Prompt B: business photography, a handcrafted ceramic bowl, earth tones, soft lighting, plants
The much more that I use Midjourney V6, the much more I’m convinced that it has no weak factors. These are the two extremely correct 3D renders of the prompt topics. I specifically like the composition of the bowl shot, with the normal light coming from the window. On the other hand, the diorama is so thorough but it does not shed that miniature feeling to it.
Individual Score: five out of five
Pastiche
Prompt A: Magneto in a dvd display grab of Dragon Ball Z, drawn by Akira Toriyama, animated by Toei animation studio, 1985 Japanese anime
Prompt B: rows on rows of lavender stretch to the horizon underneath a complete moon, in the design of vincent van gogh
I’ve in no way actually had any concern with Midjourney imitating other artists ahead of, but they’ve undoubtedly stepped up their game in V6. Their most obvious improvement is subtlety. For instance, when I created Van Gogh photographs ahead of, it did so by copying Starry Evening closely which resulted in a whole lot of spirals and stars in the sky. In V6, it took the most recognizable traits of each Van Gogh painting and designed an approximation of how the model thinks Van Gogh would paint the prompt.
Individual Score: five out of five
Architecture and Interior Style
Prompt A: interior, a shed purposed as an artwork studio, bohemian, cottagecore, normal light, whimsy, biophilic
Prompt B: exterior, a cathedral by Antoni Gaudí throughout sunset, architecture
The interior layout picture is fairly a lot excellent, in my view. The architecture shot, on the other hand, is fairly excellent itself but the intricacy of the topic led to some rendering problems. It is hidden from afar but if you zoom in, you can see some spires bleeding into each and every other.
Individual Score: four.five out of five
Text Generation
Prompt A: a bohemian coffee store named “Corner Coffee”
Prompt B: a professor creating “The Concept of Relativity” in a blackboard
Text generation continues to be a weak spot for AI picture generators, even with V6. Nevertheless, it is really worth noting that this new model may be the very best in its section for text. The corner coffee text seems a minor funky, but it is nonetheless readable for the most portion. Meanwhile, the text on the blackboard has some blunders, but you can nonetheless see what it is striving to create.
In my testing, Midjourney V6 has been extraordinary with brief texts (one-three phrases) but it gets unreadable past that.
Individual Score: four out of five
Higher Context
Prompt A: a breathtaking and cinematic portrait of a lone astronaut gazing out at the swirling nebulas of the Horsehead Nebula, their helmet reflecting the cosmic spectacle, as their massive spaceship explodes behind them. soft and dramatic lighting. evoking a sense of awe, wonder, and danger.
Prompt B: a hyper-practical portrait of an elderly female, her encounter etched with the lines of time and expertise, but her eyes shining with wisdom and warmth. she sits in a sunlit space, surrounded by mementos of a daily life effectively-lived. the portrait captures the two the elegance of age and the enduring power of the human spirit. broad-angle. inspired by rembrandt.
Midjourney is not as excellent as DALL-E three with GPT-four when it comes to prompt coherence, but it is undoubtedly up there. It missed some lines in the two prompts, like the exploding spaceship and mementos, but most of the factors are nonetheless existing, which is much more than I could say for Midjourney V5.two.
Individual Score: four out of five
Regular Score
When tallied, my regular score of Midjourney V6 is four.64 out of five. That is significantly less than half a level away from a excellent score, which exhibits how extraordinary Midjourney is at its recent stage.
If you want much more examples of Midjourney V6’s output, I hugely recommend that you read through our comparison content articles against V5 and other AI image generators.
Pros & Cons of Employing Midjourney V6
|
PROS |
CONS |
|
|
Midjourney V6 vs. Other AI Picture Generators
DALL-E three
Launched in October 2023, DALL-E three is the third model of OpenAI’s picture generator. Like V6, it was a important evolution from its preceding iteration, with a emphasis on the two comprehension and text generation. It is obtainable by way of ChatGPT Plus or with Bing Generate.
Good quality Comparison
Portraits
Landscape
3D Item Mockups
Text Generation
Higher Context Prompts
What Can make DALL-E three Greater Than Midjourney V6?
- Even now considerably far better at nuance.
- It can be accessed by way of a browser.
- Much less susceptible to AI hallucination and rendering problems.
- More rapidly generation time than the recent Midjourney model.
- GPT-four processes your conversations or prompts into ones that can be far better understood by DALL-E three.
What Can make DALL-E three Worse Than Midjourney V6?
- Midjourney can now do text far better than DALL-E three.
- Midjourney is far better at the two realism and digital artwork.
- It does not have the very same customization characteristics as Midjourney.
- You can use artist names as prompts for Midjourney.
- DALL-E does not give you manage in excess of the output’s factor ratio.
Meta
Meta’s AI picture generator is a text-to-picture generative model which makes use of a model known as Emu. It is totally totally free but it is also morally ambiguous, much more so than other picture generators, as this model makes use of information from Facebook and Instagram consumers as its education set.
Good quality Comparison
Portraits
Landscape
3D Item Mockups
Text Generation
Higher Context Prompts
What Can make Meta Greater Than Midjourney V6?
- Considerably more quickly generation velocity.
- Meta is totally free.
What Can make Meta Worse Than Midjourney V6?
- It does not conserve previous prompts and artwork.
- It does not have any customization characteristics.
- Meta’s creativity is not as excellent as Midjourney.
- Meta can not do text and does not adhere to extended prompts effectively.
- Meta makes use of Facebook and Instagram consumer information as its education set.
Wrapping Up
It is a minor as well early to inform, but if this is how excellent Midjourney V6 previously is even at its base model, then I do not see any level in investing in other AI picture generators. It is so excellent that it blows other designs out of the water. Only DALL-E can catch up to Midjourney now, and they are not even remotely shut.
That stated, it nonetheless has a couple shortcomings, specifically in comprehension and extended text generation. But then once more, so does each other AI generator.
At some level, a model will attain the level of singularity in AI picture generation, and we’ll be moving on to newer frontiers like text-to-video or picture-to-video. I really feel that Midjourney’s going to be at the pinnacle of AI picture generation — the a single that will usher this imaginative potential.
