Underused Midjourney v5 Prompt Commands :: How to use Text Weight and Image Weight

Theoretically Media
29 Mar 202311:38

TLDRIn this video, the host delves into the underused yet powerful features of Midjourney v5's prompt commands, focusing on 'Image Weight' and 'Text Weight'. The script explains how these weights can be utilized to control the output of generated images by assigning more tokens to specific keywords. It also covers the importance of prompt structure, the correct formatting for weights, and the limitations of the system. The video showcases examples of how text and image weights can be combined to achieve desired results, and discusses the use of reference images to influence the style of the output. Additionally, it explores the challenges and workarounds associated with negative prompts. The host, Tim, invites viewers to like, subscribe, and engage with the content, offering a comprehensive guide to mastering the art of Midjourney prompts.

Takeaways

  • 📝 **Understanding Weights**: In Midjourney, weights are used to control the emphasis on different parts of a prompt, with higher weights given to keywords that are more important to the final image.
  • 🔍 **Midjourney's Prompt Scanning**: The system scans prompts for keywords, assigns tokens to them, and uses these tokens to generate images, with more emphasis on words at the beginning of the prompt.
  • 🔢 **Text Weight Syntax**: To apply text weight, use a double colon followed by a number (e.g., `::1` for equal importance or `::2` for double the importance) without spaces around the colons but with a space after the number.
  • 📈 **Image Weight (IW)**: Image weight in Midjourney is used with reference images to tell the system how much to rely on the reference image for the final output, with a score between 0.5 and 2.
  • 🎨 **Combining Text and Image Weights**: You can use both text and image weights together to fine-tune the composition and style of the generated image.
  • ⚖️ **Balance in Weights**: It's recommended to keep text weights between one and ten for simplicity, but higher values can be used for more precise control.
  • 🔧 **Formatting Matters**: Incorrect formatting, such as adding spaces where they shouldn't be, can cause Midjourney to ignore the weights.
  • 🤔 **Iterative Process**: Achieving the desired image might require several attempts and adjustments to the weights and prompt order.
  • 🧩 **Photobashing**: When direct text prompts fail, techniques like photobashing (combining images) can be used to achieve the desired outcome.
  • 🚫 **Negative Prompts**: Using negative prompts (with a negative number after the double colon) can help remove unwanted elements from the image, although it can be tricky due to the system's tendency to include certain features.
  • 🔄 **Experimentation**: The process of generating images with Midjourney often involves trial and error, and playing around with different weights and prompts to get the best results.
  • 📺 **Reference Images**: Reference images can be used to inspire the style or composition of the generated image, but the output is not a direct copy of the reference.

Q & A

  • What are the two powerful techniques discussed in the video for controlling the output in mid-journey prompts?

    -The two powerful techniques discussed are Image Weight and Text Weight.

  • How does the mid-journey prompt system work when scanning for keywords?

    -Mid-journey scans the prompt for keywords, assigns tokens to those keywords, and uses those tokens to assemble the image based on its database.

  • How many tokens are typically assigned per prompt according to the video?

    -It is mentioned that approximately 75 tokens are assigned per prompt, although the exact number may have increased as the language model has been refined.

  • What is the significance of the order of keywords in a mid-journey prompt?

    -The order of keywords matters because the system places more emphasis on the words at the beginning of the prompt than those towards the end.

  • How can Text Weight be used to influence the importance of a keyword in a mid-journey prompt?

    -Text Weight is used by replacing a comma with a double colon followed by a number to indicate the relative importance of a keyword. For example, using '::1' after a keyword would give it a weight of one, while '::2' would double its importance.

  • What is the correct format for using Text Weight in a mid-journey prompt?

    -The correct format is to have no space between the keyword and the double colon, a space after the double colon and the number, and no comma following the number before the next keyword.

  • How does using Image Weight in mid-journey work?

    -Image Weight is used by adding '--IW' followed by a number between 0.5 and 2 before the prompt, which tells mid-journey how much to rely on the reference image for the final output.

  • What is the difference between image referencing and stable diffusion's posed image?

    -Image referencing in mid-journey creates an output inspired by the reference image, rather than a direct one-to-one replication. Stable diffusion's posed image would be a more direct replication.

  • How can negative prompts be used in mid-journey to remove certain elements from the generated image?

    -Negative prompts are used by adding a keyword followed by a double colon and a negative number, which instructs mid-journey to exclude those elements from the image.

  • What is the role of photobashing in the process of refining the generated images?

    -Photobashing is used when text prompts alone are not sufficient to achieve the desired result, such as removing a hat from an image. It involves manually editing the image to achieve the desired outcome.

  • What is the importance of experimenting with different weights and prompts in mid-journey?

    -Experimenting with different weights and prompts allows for greater control over the composition and elements of the generated images, helping to achieve a more accurate and desired output.

  • What advice does the video give for those who are new to using weights in mid-journey prompts?

    -The video advises starting with numbers between one and ten for Text Weight and experimenting with different Image Weight values to understand their effects on the generated images.

Outlines

00:00

🤖 Understanding Mid-Journey Prompting Techniques

The video begins with an introduction to two underutilized yet powerful techniques in mid-journey prompting: Image Weight and Text Weight. The speaker aims to clarify the confusion around these techniques by explaining how they function. Weights are used to control the output of the image by assigning more tokens to specific keywords within the prompt. The video also discusses the limitations of these techniques and offers strategies to overcome them. It's mentioned that Mid-Journey assigns around 75 tokens per prompt, with more emphasis on the beginning of the prompt. Text weights are applied by using a double colon followed by a number to indicate the importance of a keyword. The video provides examples of how different weights affect the composition of the generated images.

05:01

🖼️ Image Weighting and Reference Images in Mid-Journey

The second paragraph delves into the use of reference images in Mid-Journey and how image weights can be applied to them using the '--IW' flag followed by a number between 0.5 and 2. This indicates the level of reliance on the reference image for the final output. The video contrasts image referencing with stable diffusion's posed image and demonstrates how to achieve a balance between the reference image and the desired artistic style. Examples are given using Scarlett Johansson as Black Widow and Clint Eastwood as a character in a fictional project. The speaker also shares a trick of photobashing to combine elements from different images when Mid-Journey struggles to generate the desired output.

10:01

🚫 Experimenting with Negative Prompts in Mid-Journey

The final paragraph discusses the concept of negative prompting in Mid-Journey, which involves adding a keyword followed by a negative number to indicate elements that should be removed or de-emphasized in the output. The video shows how tricky it can be to achieve desired results with negative prompts, as the AI tends to include elements that are present in the reference image. Despite several attempts, the AI persistently includes a hat in the image, even when instructed to remove it. As a workaround, the speaker uses photobashing to manually edit out the unwanted element and suggests that this edited image could be used as a new reference for further iterations.

Mindmap

Keywords

💡Midjourney

Midjourney refers to an artificial intelligence platform that generates images based on textual prompts provided by the user. In the context of the video, it is the primary tool being discussed for creating images through the use of specific commands and techniques.

💡Text Weight

Text Weight is a technique used within the Midjourney platform that allows users to assign varying levels of importance to different keywords in their prompts. By using a colon followed by a number (e.g., '::1', '::2'), users can control how much emphasis the AI places on each element, thus influencing the final image composition.

💡Image Weight

Image Weight is another technique that involves using a reference image to guide the output of the AI. It is specified with '--IW' followed by a number between 0.5 and 2, indicating the reliance on the reference image. This helps in achieving a style or composition inspired by the reference image.

💡Prompt

A prompt in the context of Midjourney is a textual description that the AI uses to generate an image. It consists of keywords that the AI interprets and translates into visual elements. The video explains how to construct effective prompts using text and image weights.

💡Tokens

Tokens in the context of Midjourney's operation are units assigned to keywords found in a prompt. The AI uses these tokens to assemble the final image. The video mentions that there is a limit to the number of tokens assigned per prompt, which influences how the image is generated.

💡Compositional Balance

Compositional Balance refers to the visual arrangement and emphasis of different elements within an image. The video discusses how text and image weights can be used to adjust this balance, ensuring that certain elements are given more or less importance in the final output.

💡Negative Prompt

A Negative Prompt is a technique where a user specifies elements they do not want to be included in the generated image by using a negative number with the text weight syntax. The video demonstrates how this can be tricky, as the AI may still include elements that are strongly suggested by the reference image or the overall prompt.

💡Photobashing

Photobashing is a manual editing technique where parts of different images are combined to create a new image. In the video, it is used as a workaround when the AI has difficulty generating an image without certain unwanted elements, as seen when trying to remove a hat from the image.

💡Reference Image

A Reference Image is an existing image that a user uploads to serve as a visual guide for the AI when generating a new image. It helps the AI to understand the desired style or elements to include. The video shows how adjusting the image weight can change how much the output resembles the reference.

💡Illustrator Style

The term 'Illustrator Style' in the video refers to a specific visual aesthetic often associated with comic books or graphic novels. It is used to describe the desired outcome when the user is trying to get the AI to generate images with a particular artistic flair.

💡Dynamics

In the context of the video, 'Dynamics' refers to the energy, movement, or intensity portrayed in an image. The speaker uses image weight with a reference image to achieve more dynamic poses in the generated images, which adds a sense of action or liveliness to the still pictures.

Highlights

Explains the concept of Image Weight and Text Weight in mid-journey prompts

Discusses the underutilization of these powerful techniques due to confusion

Describes how mid-journey assigns tokens to keywords in a prompt

Mentions the emphasis on words at the start of a prompt over the end

Demonstrates how to use text weights with colons to control the importance of keywords

Advises on keeping text weight numbers between one and ten for simplicity

Clarifies the correct format for applying text weights without spaces or commas

Provides an example of how image composition changes with different text weights

Explains the use of reference images and image weights in mid-journey

Shows how adjusting image weight can influence the style and composition of the output

Discusses the limitations of negative prompting and the workarounds using photobashing

Illustrates the process of combining text prompts with image references for desired results

Shares a personal project example using mid-journey prompts and image weights

Provides a detailed walkthrough of experimenting with weights to achieve a specific visual style

Encourages viewers to like and subscribe for channel growth

Offers practical tips for fine-tuning prompts to get closer to the desired image

Concludes with an invitation for viewers to share questions, comments, or suggestions