[Advanced Tutorial] Midjourney Advanced Tips Guide (From Text to Image)
Midjourney is a text-to-image application similar to OpenAI's DALLE-2 and Stable Diffusion's DreamStudio. It uses some 650 million images found on the internet to generate stunning images based on the text cues provided.
Anyone who has used the Midjourney app knows that the more detailed the text provided, the more rich and unique the output image will be. Users often use raw and direct cues, such as a cowboy wearing a tuxedo on the moon, resulting in an image like this:
Beyond simple raw hint text, Midjourney offers many more options to create predictable and consistent image output! Now, let's get to the list of advanced options! Here are some of the ways you can generate image variations and the corresponding advanced settings:
1. Style - keyword (Style - keyword)
We provide a set of support prompt keywords related to "style", and you can create different output according to the style you choose. Here are some keywords and sub-genres based on the art form/design genre/artist you might wish to choose as a style: Using Design Genres as Style Cues
2. Use artist name as style
You can also specify style output by specifying an artist. Here are some sample images of the same prompts:
3. Use rendering/lighting properties as styles
4. Stylized output
You can add the setting --s (for style), and specify rendering/lighting properties as styles. The image below is from the same prompt, but with low and high stylization options.
/imagine firefighters --s 6000
Chaos - increases the level of abstraction
You can enter a number between 0 and 100 to increase or decrease the abstraction of the subject.
/imagine Eiffel tower --chaos 60
To specify the output resolution, you can use some commonly used keywords, such as 8K, 4K, photorealistic, ultra photoreal, ultra detailed, intricate details, etc. Or you can use the standard settings hd and quality/--q for predictable output. /imagine red rose flower --hd /imagine red rose flower --quality 5 By default, an image is generated with a square area (1:1 aspect ratio). But if you need a more cinematic visual, or you just want to make a laptop-friendly wallpaper, then you can change the aspect ratio of the output image. For example, if you want to imagine a picture of jasmine in a sea of flowers in the wild, you can use the following parameters to specify an aspect ratio of 4:3: /imagine jasmine in the wild flower --ar 4:3 If you need to specify the specific aspect ratio of the image Size, you can use the following parameters to set the width and height: /imagine jasmine in the wild flower --w 600 --h 300 Although you cannot directly specify a custom aspect ratio, you can use standard or non-standard ratios. . Below are some examples.
6. Graphic hints
If you want to maintain a consistent output style across multiple images, you can do so by passing the URL of the image and a text hint. For example:
/imagine http://www.imgur.com/Im3424.jpg box full of chocolates
The generated image will be based on the seed image and text hint you pass. You can also provide multiple images as hints for more detailed output.
7. Image Weight Tips
If you want the output image to be closer to the hint image you provided, you can assign a higher weight to the hint image. For example:
/imagine http://www.imgur.com/ks34f24.jpg chocolates --iw:4"--iw
The parameter is followed by a number that represents the weight of the image cue, the higher the number, the more influence that image has in the generation process.
8. Weighting of text hints
You can assign different weights to different topics by using double colons `::` in the text prompt. For example:
/imagine wild animals tiger::2 zebra::4 lions::1.5
9. Filter text hints from images
If you wish to filter out text hints when generating images, use the "--no" keyword to exclude unwanted themes. For example:
/ imagine KFC fried chicken --no sauce
This will exclude the topic "sauce" from the generated image.
10. Some interesting keywords
Here are some interesting keywords to try:
- Sony Alpha α7, ISO1900, Leica M, = specify camera type or lens type
- photorealistic, ultra photoreal, ultra detailed, complex detail, = specify detail and realism instead of art style
- unreal = specifies the Unreal Engine style
- volumetric light, cinematic lighting = specify specific lighting conditions
In summary, we have discussed how to control the type of output image to some extent, but that is not all. The amazing team at Midjourney also leaves a lot of room for your imagination and experimentation. Good luck with the different prompts and have fun creating!