PS, once regarded as a must-have skill, now seems to be in the delicate situation of "learnable but not learnable".
The latest version of Gemini's "Words to Follow", which allows you to change a picture with just one sentence, has become a phenomenal presence, and since the launch of GPT-4o's multimodal picture-generating function, the Ghibli-style pictures generated have swept the Internet almost overnight.
Facing the rise of new forces, veteran AI image generation giant Midjourney Not to be outdone, the latest version, V7, has just been officially released.
We've combed through some of the highlights of the new release:
Overall image quality is improved, text and image cues are more accurately understood, and the consistency of the body, hands, and various objects in detail is dramatically enhanced.
Character skin color and facial details more natural and clear, material texture (such as feathers, clothing) texture performance is fine, realistic light and shadow effects.
Able to better understand and present the complete scene described by the cue word, with reasonable spatial relationships between elements and a high degree of integration between background and subject.
Ability to accurately represent specific perspectives, with specialized performance in shooting styles such as low angles and close-ups.
The sense of atmosphere and emotional expression is slightly weak, lacking some drama and mystery, and the uniqueness and creativity of certain artistic styles is slightly lacking.
Note that V7 is the first version to enable model personalization by default. Users need to unlock the personalization settings to use this feature, which takes about 5 minutes and can be switched on and off at any time.
Another highlight of the V7 is Draft Mode. At half the cost of standard mode, this mode renders images up to 10 times faster.
When used on the web side, the prompt bar will automatically switch to "Dialogue Mode", and users can adjust the content directly through commands, e.g. replacing the cat with an owl or changing the scene to night, and the system will automatically adjust the prompt and generate a new image.
After clicking "Draft Mode" and enabling the microphone button, users can also enter "Voice Mode" to create in real-time through voice commands. To explicitly run a draft task, users can add the "--draft" parameter after the prompt, which is suitable for scenarios such as permutations and combinations or repetitive generation.
Midjourney officials say that 'Draft Mode' is by far the best way to iterate on ideas.
Users who are satisfied with the draft image can click the Enhance or Change buttons to re-render it in full quality. Note that the image quality in draft mode is lower than in standard mode, but the behavior and aesthetic presentation is highly consistent and suitable for quick validation of ideas.
Image from: @U79916881747113
V7 currently supports two modes of operation: Turbo and Relax. Turbo mode is faster but costs twice as much as a regular V6 job; draft mode cuts the cost in half. The standard speed mode is still being optimized and is expected to be available soon.
For now, zooming, editing and retexturing will be rolled back to the V6 model, with a gradual upgrade to V7 in the future. mood board and SREF functionality is already available, with further performance enhancements coming in subsequent updates.
The Midjourney team has revealed that new features will be released every one to two weeks for the next 60 days, with the most anticipated being the new V7 character and object referencing system.
Without further ado, let's look at the pictures.
Renowned blogger @nickfloats shared a comparison of images generated by Midjourney V6/V7 versions based on the same cue word.
Prompt: A young Indian woman with dark hair in an open ponytail and a black jacket stands on a university campus, looking directly at the camera. The image has a 1990s-style movie still aesthetic, with a close-up portrait on a sunny day. has a 1990s-style movie still aesthetic, with a close-up portrait on a sunny day.
(A young Indian woman, with black hair tied back in a loose ponytail and wearing a black jacket, stands on a college campus and looks directly into the camera. The photograph presents a close-up portrait in the style of a 90's movie stills with a sunny day in the background.)
Left: V6 Right: V7
The new version is more naturally lit, especially in the faces of the characters, with clearer skin tones, more detail and sharper focus on the image, especially in the hairstyle parts of the characters, but at the expense of some atmosphere and emotion as a result.
Prompt: A majestic barn owl perched on an ancient, moss-covered tree branch, surrounded by the misty forest. The scene is bathed in soft light filtering through the dense foliage, creating a magical and ethereal atmosphere. The scene is bathed in soft light filtering through the dense foliage, creating a magical and ethereal atmosphere. Photorealistic style with attention to detail of the feathers and textures.
(A majestic barn owl perches on an ancient, moss-covered branch surrounded by a misty forest. Soft light pours through the dense foliage, creating a mysterious and ethereal atmosphere. The style of the image is realistically realistic, with detailed textures of feathers and bark.)
Left: V6 Right: V7
The new version of the owl's feather texture is more delicate, with stronger layers of feathers and light and shadow effects, making it appear more realistic. As for the posture of the owl, the new version is more natural, the body is slightly tilted to appear more dynamic, and the details of the eyes are more vivid, conveying a vigilant demeanor.
In short, you can't have it both ways. The V7 is better at pursuing the goal of realism; the V6 may be better at pursuing image impact and mystery.
Prompt: A person's hand points towards the window of an airplane, which is seen from inside with its wing visible in profile. The sky outside shows a clear horizon at dawn or dusk. In front of them lies a vast expanse of ocean.
(A man's hand points out the airplane window, which shows the wings of the plane visible from the side. Outside, the sky is clear and the skyline shows dawn or dusk. A vast ocean is in front of the eye.)
Left: V6 Right: V7
The same is the scene from the airplane porthole to look out, one hand to the window, V7's progress can be seen with the naked eye, by adding the elements of the aircraft wing, increasing the level of the picture and the sense of realism, more feel the perspective of being in the aircraft.
Prompt: A close-up of broccoli being sautéed in oil, with the vegetable's green florets contrasting against a dark brown sauce. A large spoon is partially visible inside a stainless steel pot filled with caramelized and shiny black glaze on top, against a white... visible inside a stainless steel pot filled with caramelized and shiny black glaze on top, against a white...
(A close up shot of a piece of broccoli tossed in oil, the lush green cauliflower contrasting with the dark brown sauce. A large soup ladle is partially visible, resting in a stainless steel pan covered in a caramel-colored glossy black sauce with a white background...)
Left: V6 Right: V7
V6 While the focus is clear with the close-up shots highlighting the sauce and the spoon, and attention is easily drawn to the texture and detail of the sauce, the new version presents a more complete cooking scenario that logically fits better with the actual cooking process if you look at the requirements of the cue word by showing the entire pan of broccoli and the sauce with it.
Prompt: Close-up of an anime woman's face with a shocked expression, dark hair, in the anime style. Colorful animation stills, close-up intensity, soft Colorful animation stills, close-up intensity, soft lighting, low-angle camera view, and high detail.
(Close-up of an anime woman's face with shocked expression and black hair, rendered in anime style. Colorful animated image with strong close-ups, soft lighting, low angle shot with very fine details.)
Left: V6 Right: V7
In terms of lighting and color handling, V7 also meets the "SOFT LIGHTING" requirement in the tip word, with softer and more natural transitions of highlights and shadows on the face, especially the distribution of light and shadow on the eyes and cheeks, creating a more three-dimensional effect.
Prompt: A dynamic film still of an epic space battle with sleek starfighters zooming past a massive space station, lasers firing, and a distant planet visible in the backdrop. visible in the backdrop.
(A dynamic cinematic picture of an epic space battle, with streamlined starfighters whizzing by, giant space stations on the sidelines, lasers firing in all directions, and planets clearly visible in the distance, makes for a stunning picture.)
Left: V6 Right: V7
For portraits, the v7 improves on the fineness of detail and realism (clothing, skin, lighting, etc.), with more three-dimensionality and blending of the characters with the background, but regresses on emotional delivery, drama of expression, and dynamism.
Below are three specific examples of comparisons:
Prompt: 1980s mystery film, low-angle shot of an evil-eyed French Butler sporting a black suit and grasping a candle in the hallway of a creepy Victorian mansion with musty decor. The warm candle glow evokes a spooky sense of mystery
(A low angle shot in the style of a 1980s mystery movie captures an evil-eyed French butler, dressed in a black suit and holding a candle in his hand, standing in the hallway of a musty Victorian mansion with dated decor. The warm candlelight creates an eerie, mysterious atmosphere.)
Left: V6 Right: V7
Prompt: 1990s medium-full street style fashion photo shot on Kodak 500T capturing a rugged 50-year-old man with curly gray hair, 5-o'clock shadow, and a stern look walking down the sidewalk on a bright spring morning in Paris. He's wearing ...
(A 1990s medium full-body street style photograph, shot on Kodak 500T film, capturing a man in his fifties with curly, grayish-white hair, stubble, and a serious expression as he walks down a sidewalk on a bright spring morning in Paris. He is wearing...)
Left: V6 Right: V7
Prompt: Cinematic, off-center, two-shot, 35mm film still of a 30-year-old French man, curly brown hair and a stained beige polo sweater, reading a book to his adorable 5-year-old daughter, wearing fuzzy pink pajamas, sitting in a cozy corner nook...
(A cinematic 35mm film image, composed off-center, captures a 30-year-old French man with curly brown hair and a soiled beige polo shirt reading to his adorable 5-year-old daughter.) The daughter, wearing soft pink pajamas, sits in a cozy corner.)
Left: V6 Right: V7
Will Conch AI + Midjourney be the king combo to get pictures moving? Netizen @ainextastro gave it a try too.
The photo below, generated by @tanvitabs using v7, goes over all the pitfalls of AI image generation, including an extra third hand out of nowhere, confusion between t-shirt and suit attire, and mismatched face generation.
Here are four photos generated based on the same cue word, which version do you prefer?
Prompt：Elegant female model standing by a large window in a sunlit room, soft morningliaht casting natural highlights and shadows on her face, wearing a flowing beigedress, relaxed pose, minimal makeup, 85mm lens depth of field, lifestyle fashioneditorial, cinematic tones, airy atmosphere
(An elegant female model stands next to a large window in a sun-drenched room, the soft morning light casting natural highlights and shadows on her face, wearing a long, flowing beige dress, a relaxed pose, and minimalist makeup.) Shot with an 85mm lens, the background has a shallow depth of field, creating a lifestyle fashion editorial feel, with cinematic tones and an airy feel to the image).
Over the past few years, the main theme of AI image generation has been the quest for realism and greaselessness.
Last year, Flux made a name for itself by generating hyper-realistic portraits.Now, Midjourney V7 takes over the baton and once again pushes 'reality' to new heights with richer details, more layered light and shadow, and more natural skin texture.
But is the pursuit of truth, the end of all?
With the recent release of the GPT-4o multimodal image generation feature, Ghibli-style images lit up the web almost overnight and blew into the world of AI image generation like a spring breeze.
There are more possibilities for technological choices, and the creative playground has changed along with them. To put it in more colloquial terms, it's now like standing in the kitchen, whether you want to make a down-to-earth meal or something a little darker, it's all up to you.
