Article directory
Wen Xin Yi Yan supports continuous dialogue
I'm just messing around and I'm not a professional haha
first step
Hello, today we are going to create an image using a diffusion model. I'll give you some information. OK?
Step 2
Here’s how Midjourney works: Midjourney is another AI-based tool that generates images based on user prompts. MidJourney excels at adjusting the actual art style to create any combination of images the user desires. It excels at creating environments, especially fantasy and sci-fi scenes, with dramatic lighting that looks like concept art for video games.
How does Midjourney work? Midjourney is an AI image generation tool that takes text prompts and parameter input and uses machine learning (ML) algorithms trained on large amounts of image data to generate unique images. Powered by Latent Diffusion Model (LDM), a cutting-edge text-to-image synthesis technology. Before understanding how ldm works, let's first look at what a diffusion model is and why we need ldm. The Diffusion Model (DM) is a transformer-based generative model that takes a piece of data, such as an image, and gradually adds noise over time until it becomes unrecognizable. From there, they try to reconstruct the image into its original form, learning how to generate the image or other data in the process. The problem with DM is that powerful DM often consumes hundreds of GPU days and inference is very expensive due to sequential computation. To enable DM to be trained on limited computing resources without compromising its quality and flexibility, DM is applied to the latent space of powerful pre-trained autoencoders. Training a diffusion model on this representation can hit a sweet spot between complexity reduction and detail preservation, significantly improving visual fidelity. Introducing cross-attention layers into the model architecture turns the diffusion model into a powerful and flexible generator for general conditional inputs such as text and bounding boxes, enabling convolution-based high-resolution synthesis. Wait, I have more information to provide.
third step
Version Light Midjourney regularly releases new model versions to improve efficiency, consistency and quality. The latest model is the default, but other models can be used. Different models are good at different types of images. The Midjourney V5 model is the latest and most advanced model, released on March 15, 2023. To use this model, add the -v 5 parameter at the end of the prompt, or use the /settings command and select MJ Version 5. This model is very consistent, excels at interpreting natural language cues, has higher resolution, and supports advanced features such as tile repeat patterns. Open type -v 5 after the prompt or select "V5" from /settings What's new in the V5 base model? Wider range of styles, more responsive to prompts, higher image quality (2x higher resolution) Improved Increased dynamic range, more detailed images. Details are more likely to be correct. Reduce unnecessary text. Improved the performance of image prompts, supports seamless tile tile parameters (experimental), supports aspect ratios greater than 2:1 (experimental), supports iw, which is used to weigh image prompts and text prompt styles and V5 prompts.
Today's test was basically a "Pro" mode model.
It is more "unbiased" than v3 and v4, and is tuned to provide wide output diversity and is very sensitive to your input. -The trade-off here is that it may be harder to use. Short prompts may not work well. You should try writing longer, more specific words that describe what you want (eg: "cinematic photos with dramatic lighting").
Please chat with each other in the prompt chat to learn how to use v5.
We would like to have a "friendly" default style in v5 and then switch to the default style later. When that happens, we'll still let you turn it off and go back to "original" mode for today. Please note this is an alpha test and things will change. Do not rely on this exact model being available in the future. When we release V5 to the full version, it will be significantly revised.
There is currently no V5 upsampler, and V5's default resolution is the same as the upgraded V4. If you click "High" it will immediately give you a picture. Community Standard: This model produces more realistic images than anything we've released before.
We have increased the number of moderators, improved moderator tools, and will enforce our community standards more rigorously and rigorously. Don’t be a jerk and don’t create drama. More information about V5: V5 is the second model we have trained on the AI supercluster and has been working for 5 months. It uses significantly different neural structures and new aesthetic techniques. V5 is not the last step, but we hope you all feel the progression of something deep and unfathomable in our collective human imagination. Wait, I have more information to provide.
the fourth step
Basic parameters aspect ratio -Aspect, or -ar change the generated aspect ratio. Chaos—Chaos <number 0 - 100> changes the degree of variation in the results. Higher values will produce more unusual and unexpected generations. There is no—no negative cues—no plants trying to remove plants from the image. Quality—Quality<. 25, .5, 1, or 2>, or -q <. 25, 0.5, 1 or 2> how much render quality time you want to spend. The default value is 1. Higher values cost more, lower values cost less. Seed — Seed < an integer between 0-4294967295 > The Midjourney bot uses the seed number to create a visual noise field, like TV static, as a starting point for generating an initial grid of images. The seed number is randomly generated for each image, but can be specified with the --Seed or --sameseed parameters. Using the same seed number and prompt will produce similar ending images. Stop—Stop <an integer between 10 and 100> Use the --Stop parameter to complete the job in the middle of the process. Stopping the job at an earlier percentage may produce blurry, less detailed results.
Style - Style <4a, 4b or 4c> Switches between versions of the Midjourney model version 4 Stylize - The Stylize or -s parameter affects how much Midjourney's default aesthetic style is applied to Jobs. When the U button is selected, another "light" upgrader is used. The result is closer to the original grid image. The upgraded image has less detail and is smoother. When the U button is selected, an optional beta upgrader is used. The result is closer to the original mesh image. The upscaled image adds significantly less detail. Default (Model Version 5) Aspect Ratio Chaos Mass Seed Stop Style Stylized Default 1:1 0 1 Random 100 4c 100 Range Any 0 - 100 .25 .5 1 or 2 integers 0 - 4294967295 10 - 100 - 0 - 1000 Aspect ratios greater than 2:1 are experimental and may produce unpredictable results.
Compatibility Model Version & Parameter Compatibility Impact Initial Generation Impact Change + Remix Version 5 Version 4 Version 3 Test/TestpNiji Max Aspect Ratio ✓✓ 1:2 or 2:1 5:2 2:5 3:2 or 2:3 1:2 or 2:1 Mess ✓✓✓✓✓✓Image Weight✓✓✓✓No✓✓✓✓✓✓✓Quality✓✓✓✓✓Seeds✓✓✓✓✓✓Sameseed✓✓Stop✓✓✓✓✓ ✓✓ Style 4 a and 4 b stylized ✓ Default 0 – 1000 = 100 0 – 1000 Default = 100 625 – 60000 Default = 2500) 1250 – 5000 Default = 2500) Tiles ✓✓✓✓Video ✓✓ Number of grids Pictures - - 4 4 4 2 (1 when aspect ratio ≠ 1:1) But wait, I have more information to provide.
the fifth step
Okay, now I'm going to give you some examples of hints used in Midjourney V5. OK?
Step 6
Prompt 1: Super wide angle, modern photo of Hawaiian beauties in the 1970s. This photo was taken by Mary Shelley with a Nikon D5100 camera, using aperture off/2.8, ISO 800, and a shutter speed of 1/100 second. UHD dtm HDR 8k --ar 2:3 --v 5
Prompt 2: A steampunk-inspired, futuristic battle-ready motorboat skims the water with a fierce presence. Intricate gears and brass fittings adorn its hull, showcasing the perfect combination of advanced technology and Victorian aesthetics. This masterpiece of realism gleams in the sun and is ready for action. --ar 16:10 --s 50 --v 5 --q 2
Prompt 3: Epic background art, simple hacker theme, divine color scheme, cryptic codes, alphanumeric sequences, magic, high quality 4k, render value -v 5 -ar 9:16
Prompt 5: Full body blonde beauty, wearing brown jacket, photography, Canon EOS 5D Mark IV SLR camera, EF 50mm f/1.8 STM lens, resolution 30.4 million pixels, ISO sensitivity: 32000, shutter speed 8000 seconds-- - 9: 16 - -Zoom-- -v 5.
Prompt 6:: Hasselblad 24mm full-body photography, gorgeous and satisfied African women, delicate and natural skin, no makeup, delicate eyes, long braids – ar2:3–q5–v5–v4.
Prompt 7: Beautiful dark red sunset at night by the sea, complex, stunning, beautiful, realistic, super high resolution, wide angle, depth of field, π dynamic lighting -ar 1:2 -v 5
Can you now understand how the prompt word "Midjourney" is formed? Yes or No
Step 7
Very good. Here are some more examples of Midjourney prompts.
Prompt 1: Hasselblad 24mm full-body photography, gorgeous and satisfied African women, delicate and natural skin, no makeup, delicate eyes, long braids –ar 2:3 --q 5 --v 5 --v 4.
Prompt 2: Beautiful dark red sunset at night by the sea, complex, stunning, beautiful, realistic, super high resolution, wide angle, depth of field, dynamic lighting -ar 1:2 -v 5
Prompt 3: A stunning, hyper-realistic photo of a ferocious Viking warrior meticulously sharpening his powerful blade in the wilds of the rugged, untamed Scandinavian landscape. This scene was captured with a Nikon D850 camera using a 70-200mm f/2.8 lens, highlighting every intricate detail of the Viking's weathered face, war-worn armor and the expert craftsmanship of his weaponry. The settings used were aperture closed/4, ISO 400, shutter speed 1/200 second, balancing natural light and shadow to emphasize the intensity and determination in the Viking eyes. Juxtaposing the raw power of the warrior with the serene beauty of the surrounding environment, this composition captures the essence of the Viking spirit in stunning high-resolution imagery, transporting viewers back to legendary battles and untold stories. story. –ar 16:9 --q 1.5 --v 5.
Prompt 4: A stunning and atmospheric 1970s New York street cafe captures a nostalgic and cinematic style reminiscent of the golden age of cinematography. This retro scene showcases bustling city life, with customers enjoying coffee at outdoor tables, surrounded by classic cars and retro architecture. This photo was cleverly composed using a Leica M3 rangefinder camera paired with a Summicron 35mm f/2 lens, renowned for its clarity and beautiful color rendering. The photo was shot on Kodak Portra 400 film, giving it a warm and timeless color palette that enhances the overall atmosphere. The photographer cleverly used a shallow depth of field and an aperture of off/2.8 to isolate the cafe and its patrons from the bustling city background. The ISO was set to 400 and the shutter speed was 1/125 second, capturing the perfect balance of light and movement. Soft, diffuse sunlight filters through the iconic New York skyline, casting warm golden tones across the scene and highlighting the rich textures of the brick buildings and cobblestone streets, further enhancing the composition. –ar 3:2 --q 2.
Prompt 5: Pov high definition macro photography of a realistic cat wearing reflective sunglasses relaxing on a tropical island, dramatic light - 2:3 -s 750 -v 5 Thanks for the example tips for use in Midjourney V5 . These tips are a good example of how detailed and specific text tips can be for producing images with desired characteristics. These tips also show the usage of various parameters such as aspect ratio, stylization, version and quality settings. These examples will help understand how to use Midjourney V5 to create effective prompts for generated images.
Can you now understand how the prompt word "Midjourney" is formed? Yes or No
Step 8
Very good. Now I want you to play a professional photographer. When describing photo prompts, you'll use rich, descriptive language, including camera settings. Now the first prompt I want you to create is a photo of a female influencer from the 1930s. Get inspired by the formatting of the example prompts, don't copy them but use the same formatting. The content of the prompt word should be limited to 399 words.
Haha, copy and paste the prompt word into the AI painting
