Midjourney Prompt Tips: Codes & Tricks - WHAT STILL WORKS? (Perfect for AI Art Beginners!)

Wade McMaster - Creator Impact
16 Feb 202321:59

TLDRThis video offers valuable tips for AI art beginners using Midjourney, explaining which command codes still work after the release of version 4. It covers aspect ratio control, chaos for variance, negative words, stop for partial rendering, and seed for reproducibility. The script also clarifies which codes are algorithm-specific, such as 'same seed', 'style', 'image weight', 'video', 'quality', 'stylize', 'creative', and 'tile'. Additionally, it discusses the various upscaling options available and how to set custom commands for convenience, aiming to enhance control over AI-generated art.

Takeaways

  • 😀 Midjourney has released version 4, which has updated algorithms and some codes no longer work as they did before.
  • 🔍 Users can check available algorithms by going to Discord's settings to see the versions ranging from one to four and niji mode for anime-style images.
  • 📏 The aspect ratio (AR) code works across all algorithms, allowing control over image dimensions like 1:1 or 3:2.
  • 🔄 The chaos code introduces variance among images, with higher values like 100 creating more diverse results.
  • 🚫 The 'no' code helps to de-emphasize certain elements in the generated images, providing more control over the final output.
  • 🛑 The 'stop' code lets users halt the rendering process at a specific percentage, creating unique partial images.
  • 🌱 The 'seed' code allows for reproducing the same image by using the same starting point for the algorithm.
  • 🌐 The 'same seed' code works with versions one to three and test versions, ensuring consistency across images.
  • 🎨 The 'style' code with version 4 lets users choose between 4A, 4B, or 4C for different visual outcomes.
  • 🖼️ The 'image weight' (IW) code increases the influence of a provided image URL on the generated result, and it's only compatible with versions up to three and test versions.
  • 🎬 The 'video' code creates short video clips of the generated images, but it's not supported in the latest version or niji mode.
  • 🔍 The 'quality' setting adjusts the rendering time and detail level of the images, with higher values like 2 or 5 resulting in more detailed outputs.
  • 🎨 The 'stylize' code, with different ranges for each version, allows for more artistic freedom in the interpretation of the prompt.
  • 🛠️ The 'creative' code boosts the creativity of the test algorithms, adding more color and vibrancy to the images.
  • 🔄 The 'tile' code creates images that can be tiled seamlessly, useful for pattern creation in design work.
  • 🔧 Custom commands and suffixes can be set using the 'prefer' command in Midjourney for quicker access to frequently used options.
  • 🔄 Multi-props and prompt weight allow for more nuanced control over the interpretation of prompts, emphasizing or de-emphasizing certain words.

Q & A

  • What is the purpose of the aspect ratio (AR) code in Midjourney Prompts?

    -The aspect ratio (AR) code in Midjourney Prompts is used to control the proportions of the generated image. By specifying a ratio like 1:1 for square images or 3:2 for a more portrait-oriented image, users can dictate the shape of the output according to their needs.

  • How does the 'chaos' code affect the variation in generated images?

    -The 'chaos' code introduces variance among the images in a grid. A chaos level of 0 results in very similar images, while increasing the chaos level to 100 or more creates a greater diversity in the image results, offering more unique outcomes in each grid.

  • What is the function of the 'no' code in Midjourney Prompts?

    -The 'no' code is used to exclude certain elements from the generated images. By adding 'no' followed by specific words, those words are de-emphasized or omitted from the images, giving users more control over the content of their prompts.

  • Can you explain the 'stop' code and its effect on image rendering?

    -The 'stop' code allows users to control the completion level of the image rendering process. By specifying a percentage, such as 'stop at 10', the rendering process stops at 10% completion, creating an unfinished look. This can be used creatively to achieve specific visual effects.

  • What is the role of the 'seed' code in generating consistent images?

    -The 'seed' code is used to replicate the same starting point for image generation. By using the same seed number, the algorithm produces images that are very similar to the original, allowing for consistent results when re-entering a prompt.

  • Why might the 'same seed' code not work as expected with test algorithms?

    -The 'same seed' code may not produce as consistent results with test algorithms because these algorithms might handle the seed differently compared to versions one, two, and three. This inconsistency can lead to minor differences in the generated images.

  • How does the 'style' code differ between version 4 and other versions of Midjourney?

    -The 'style' code in version 4 allows users to select between 4A, 4B, and 4C, with 4C being the default. Each style option results in different visual outcomes, with 4A being more colorful and illustrative, while 4B and 4C offer variations in photorealism and drama.

  • What is the purpose of the 'image weight' (IW) code and which versions does it work with?

    -The 'image weight' (IW) code increases the influence of a provided image URL on the generated output. It works with versions one to three and the test algorithms but not with version four or niji. By assigning a weight from one to five, users can control the degree to which the algorithm relies on the reference image.

  • Can you describe the 'video' code and its compatibility with different Midjourney versions?

    -The 'video' code is used to generate a video based on the prompt. It is compatible with versions one, two, three, and the test algorithms but does not work with the current default version four or niji. When used, it creates an MP4 file that can be downloaded or viewed directly.

  • What does the 'quality' code do and on which versions is it effective?

    -The 'quality' code adjusts the rendering time and detail level of the generated images. It is effective on versions one, two, three, four, and niji. Lower values like 0.25 render images faster with less detail, while higher values up to five take longer to render, resulting in more detailed images.

  • How does the 'stylize' code affect the interpretation of prompts?

    -The 'stylize' code allows the algorithm to interpret prompts with more freedom, resulting in more stylized and varied images. It works on all algorithms except niji, with different ranges for version 4 (0 to 1000), versions one to three (625 to 60000), and test algorithms (1250 to 5000).

  • What is the 'creative' code and which algorithms does it work with?

    -The 'creative' code enhances the creativity of the generated images, adding extra color and vibrancy. It only works with the test and test P algorithms, providing a unique way to stylize and interpret prompts differently.

  • Can you explain the 'tile' code and its benefits for graphic designers?

    -The 'tile' code creates images that can be tiled repetitively without visible seams, making it ideal for pattern creation. It works with versions three and the test algorithm. Graphic designers can use this feature to create seamless patterns and backgrounds.

  • What are the different upscaling options available in Midjourney, and how do they differ?

    -Midjourney offers several upscaling options, including regular upscale, light upscale, beta upscale, and max upscale. Each option increases the resolution differently, with the max upscale providing the highest resolution of 16384 by 16384 pixels. The light upscale reduces detail, while the beta upscale increases resolution without adding much detail.

  • How can users customize their Midjourney experience with 'prefer' commands?

    -Users can customize their Midjourney experience using 'prefer' commands such as 'prefer suffix' to automatically add certain options to prompts, 'prefer option set' to create custom codes with specific settings, and 'multi-props' to separate words in prompts for different interpretations.

  • What is prompt weight or word weight, and how can it be used in Midjourney Prompts?

    -Prompt weight or word weight allows users to emphasize certain words in their prompts over others by assigning numerical values from one to five, or zero. This can change the interpretation of the prompt, making some elements more or less prominent in the generated images.

Outlines

00:00

🛠️ Controlling Art Generation with Codes in Mid Journey

This paragraph discusses the changes in code functionality with the release of Mid Journey version 4. It explains that not all codes work with the new version and some algorithms, and proceeds to demonstrate which codes are effective. The speaker guides viewers on how to navigate Discord settings to understand different algorithms, including versions 1-4, niji mode for anime-style images, and MJ test algorithms for high-resolution outputs. The focus is on aspect ratio control, chaos for variance, negative words to exclude elements, and the stop code to control render completion. Each code's compatibility and functionality across different algorithms are detailed.

05:01

🔄 Exploring Advanced Coding Features in Mid Journey

The speaker delves into advanced coding options for Mid Journey, starting with the seed code for replicable results and same seed for consistent image generation across different versions. It discusses the style code's impact on photorealism and drama, and the image weight (IW) code's influence on image priority in generation. The video code for creating videos from prompts is introduced, along with its limitations to certain versions. The paragraph concludes with the quality code's role in adjusting render time and detail, and the stylize code's effect on image interpretation freedom, with specific ranges for different versions.

10:01

📹 Utilizing Video and Upscaling Features in Mid Journey

This section covers the video feature's creation of MP4 files from image prompts, noting potential bugs and the process of obtaining upscaled images. It explains the quality command's application across various versions and the niji algorithm, allowing for faster or more detailed rendering. The stylize command's exclusion in the niji algorithm is mentioned, along with its varying ranges for different versions. The creative command's enhancement of image vibrancy in test algorithms is showcased, followed by the tile command's ability to create seamless, tiling images, particularly useful for graphic design and pattern creation.

15:01

🔍 Customizing Mid Journey with Upscaling and Settings

The paragraph explains the upscaling process in Mid Journey, detailing the default 512x512 pixel grid size and the options for different resolutions depending on the algorithm version. It describes the settings accessible through the '/settings' command, allowing customization of default upscale types and the introduction of beta and light upscale options. The paragraph also touches on the max upscale feature for versions one to three, the remastered option, and the documentation for understanding the different upscale results. The mention of Topaz Gigapixel AI for superior image enlargement concludes the discussion on upscaling.

20:02

📝 Custom Commands and Multi-Props in Mid Journey

The final paragraph introduces custom command settings in Mid Journey, such as 'prefer suffix' for automatic addition of aspect ratios and 'prefer option set' for creating named command sets. It demonstrates how to list and remove custom options. The multi-props feature is explained, allowing for the separation of words in prompts to change the interpretation, and the introduction of prompt or word weight to emphasize or de-emphasize certain words in a prompt. The speaker wraps up by summarizing the video's content on useful codes and commands for Mid Journey and invites viewers to check out more content on their channel.

Mindmap

Keywords

Aspect Ratio

Aspect ratio refers to the proportional relationship between the width and height of an image or screen, commonly expressed by two numbers separated by a colon. In the context of the video, the aspect ratio is used to control the dimensions of the generated AI art, such as '1:1' for a square image or '3:2' for a more rectangular shape. The script mentions that different algorithms have limitations on the aspect ratios they can produce.

Chaos

In the video, 'chaos' is a parameter used in AI art generation to introduce variability among the images produced from the same prompt. A chaos value of zero results in images that are nearly identical, while a higher value like 100 creates more diverse outputs. It's a way to explore different interpretations of the same concept without altering the original prompt.

Negative Words

Negative words, denoted by 'no' in the script, are used to exclude certain elements from the AI-generated images. By specifying 'no' followed by a word, the AI is instructed to de-emphasize or avoid including that word in the final artwork. This feature is showcased in the script as a way to refine the content of the generated images.

Stop

The 'stop' code in the video is a directive to halt the rendering process at a certain percentage. For instance, using 'stop at 10' will result in an image that is only 10% complete. This feature can be used to create unique and abstract visual effects, as demonstrated by the script.

Seed

A 'seed' in the context of AI art generation is a numerical value that initiates the random number generation process, which influences the final output. By using the same seed value with a given prompt, the AI can produce similar or identical images, as the script illustrates by copying a seed from one image and using it to recreate a similar result.

Same Seed

As mentioned in the script, 'same seed' is a command used to ensure that all images in a set are generated with the same initial conditions, leading to a consistent look across the board. This is useful for creating image sets where uniformity is desired.

Style

The 'style' code in the script refers to different modes within the AI algorithm that can affect the artistic interpretation of the prompt. For version 4 of the algorithm, styles 4A, 4B, and 4C are mentioned, each producing distinct visual outcomes, with 4C being the default setting.

Image Weight

Image weight, indicated by 'IW' in the script, is a parameter that assigns a level of importance to a reference image when generating new artwork. A higher image weight means the AI will rely more heavily on the reference image for guidance, as shown when the script compares an image with an image weight of one versus five.

Video

The 'video' command in the script is used to generate a video file alongside the static image. This feature is available for certain versions of the algorithm, and the script demonstrates how to access the video file by reacting with an envelope after image generation.

Quality

In the context of the video, 'quality' is a setting that adjusts the rendering time and detail of the AI-generated images. A lower quality setting like 0.25 will produce images faster, while a higher setting like 2 or 5 will take more time to render, resulting in more detailed and refined images.

Stylize

Stylize is a parameter that allows the AI to interpret the prompt with more artistic freedom, as opposed to a literal interpretation. The script explains that different versions of the algorithm have different ranges for the stylize value, and adjusting this value can lead to more varied and creative outputs.

Creative

The 'creative' command, as discussed in the script, is specific to the test and test P models and is used to enhance the AI's creative interpretation of the prompt. It results in more vibrant and stylized images, as exemplified by the script's comparison of a standard image versus one generated with the creative command.

Tile

The 'tile' command in the script is used to create images that can be repeated seamlessly to form patterns. This feature is particularly useful for graphic designers or those creating backgrounds and textures, as it allows for the generation of large, continuous patterns from a single image.

Upscaling

Upscaling in the video refers to the process of increasing the resolution of an image while maintaining or enhancing its quality. The script outlines different upscaling options available for various versions of the algorithm, including standard, light, beta, and max upscaling, each producing images of different sizes and levels of detail.

Multi-props

Multi-props, as explained in the script, is a feature that allows for the separation of words in a prompt to be interpreted individually rather than as a single phrase. This can lead to unique interpretations of the prompt, as demonstrated by the script's example of separating 'hot dog' into 'hot' and 'dog'.

Prompt Weight

Prompt weight, also known as word weight, is a technique mentioned in the script where numbers are assigned to words in a prompt to indicate their relative importance. By giving a word a higher number, such as 'hot' with a weight of 5 and 'dog' with a weight of 1, the AI will emphasize the more important word in the image generation process.

Highlights

Midjourney Prompt Tips discusses the effectiveness of various codes and tricks for AI art generation.

Version 4 of the AI has been released with some codes no longer functioning as they did before.

The 'aspect ratio' code (e.g., --AR 1:1) is functional across all algorithms to control image dimensions.

The 'chaos' code introduces variance in image grids, with values from 0 to 100 affecting the diversity of results.

The '--no' code can be used to de-emphasize certain words in prompts, providing more control over the generated images.

The '--stop' code allows for stopping the rendering process at a specified percentage for unique image effects.

The '--seed' code enables the reproduction of similar images by using the same starting point for the AI's noise pattern.

The 'same seed' code works with versions 1-3 and test algorithms but may not be as consistent with test and test P.

The '--style' code with 4A, 4B, or 4C options adjusts the style of images in version 4, with 4C being the default.

The '--IW' or 'image weight' code prioritizes the importance of an image URL in the generation process, not available in version 4 or niji.

The '--video' code creates videos from image prompts, compatible with versions 1-3 and test algorithms but not with the default version 4.

The 'quality' setting adjusts rendering time and image detail, with values from 0.25 for faster rendering to 5 for highest quality.

The '--stylize' code, with different ranges for various versions, allows for more freedom in AI's interpretation of prompts.

The '--creative' code boosts creativity in the test and test P algorithms, adding color and vibrancy to images.

The '--tile' code creates images that can be tiled seamlessly, useful for graphic designers and pattern creation.

Upscaling options vary by algorithm version, with test algorithms producing higher resolution images by default.

Custom commands and suffixes can be set using '/slash settings' and 'prefer suffix' for streamlined prompt creation.

Multi-props and prompt weight allow for greater control over the interpretation of prompts by emphasizing or de-emphasizing certain words.

The video concludes with a summary of the most useful codes and commands for mid-journey with current and old algorithms.