Underused Midjourney v5 Prompt Commands :: How to use Text Weight and Image Weight
TLDRIn this informative video, the host delves into the underutilized yet powerful features of Midjourney v5, specifically focusing on Image Weight and Text Weight. The host explains how these weights can be used to control the output of the AI-generated images by assigning more tokens to specific keywords within the prompt. The video showcases examples of how changing the weights can significantly alter the composition of the final image. Additionally, the host discusses the limitations of these features and provides tips on how to work around them. The video also touches on the use of reference images and negative prompts to further refine the image generation process. The host's practical demonstrations and detailed explanations offer viewers a comprehensive understanding of how to effectively use these features to create more precise and desired outputs.
Takeaways
- 📝 **Understanding Weights**: Midjourney uses tokens to assemble images based on keywords, with more emphasis on words at the start of the prompt.
- 🔢 **Text Weight**: You can control the importance of keywords by using a colon colon followed by a number, which allocates more tokens to that keyword.
- 🔄 **Combining Weights**: Text and image weights can be combined to control the output composition more precisely.
- 🚫 **Limitations**: There are limitations to how weights work, and sometimes it takes a few attempts to get the desired image.
- 📐 **Aspect Ratio**: Midjourney considers the aspect ratio when assembling images, which can affect the final composition.
- 🔑 **Formatting**: Correct formatting is crucial for weights to be recognized; no space between the keyword and the colon colon, but there is a space after the colon colon to the number.
- 🍰 **Example Usage**: For instance, 'cupcake' can be weighted to focus more on the 'cake' part, resulting in an image that emphasizes the cake.
- 🖼️ **Image Weight (IW)**: Reference images can be used with an image weight (IW) between 0.5 and 2 to influence the style and composition of the output.
- 🎨 **Stylistic Influence**: Image weights can help achieve a specific artistic style, like mimicking the work of a comic book artist.
- 🔍 **Fine-Tuning**: Weights allow for fine-tuning of image composition, especially useful in longer prompts with multiple elements.
- ⛔️ **Negative Prompts**: Negative prompts can be used to remove unwanted elements from the image, although they can be tricky to apply effectively.
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, which allow users to control the composition and emphasis of elements in the generated images.
How does the mid-journey system assign tokens to keywords in a prompt?
-Mid-journey scans the prompt for keywords and assigns tokens to these keywords. The tokens are then used to assemble the image based on the database of the mid-journey system. It's mentioned that about 75 tokens are assigned per prompt, with more emphasis on the words at the start of the prompt.
What is the purpose of using text weights in mid-journey prompts?
-Text weights are used to add more tokens to specific keywords, thus increasing their importance in the image generation process. This allows the user to control which elements of the prompt are emphasized in the final image.
How do you format a text weight in a mid-journey prompt?
-Text weights are formatted by using a double colon (::) after the keyword, followed by a number that represents the weight. For example, 'keyword::2' would give that keyword twice the importance of a keyword without a weight.
What is the role of image weight in mid-journey prompts?
-Image weight is used when a reference image is provided. It scores the importance of the reference image in the final output, with a range from 0.5 to 2, telling mid-journey how much to rely on the reference image.
How does the mid-journey system handle negative prompts?
-Negative prompts are used to remove certain elements or features from the generated image. They are formatted by adding a keyword followed by a double colon and a negative number, such as 'keyword::-1'.
What is the significance of the order of keywords in a mid-journey prompt?
-The order of keywords in a prompt is significant because the system places more emphasis on the words at the beginning of the prompt compared to those at the end.
What does the video suggest as a good practice when using text weights?
-The video suggests using numbers between one and ten for text weights for simplicity and ease of calculation, although higher numbers are possible if the user is comfortable with more complex calculations.
How does the video demonstrate the use of text weights?
-The video demonstrates the use of text weights by showing how changing the weights of keywords in a prompt, such as 'cupcake' and 'coffee', can influence the focus of the generated image.
What is the recommended approach when dealing with complex prompts that require fine-tuning?
-The video suggests using natural language prompts for simpler requests, but for complex prompts with multiple elements, adjusting the text and image weights can help achieve the desired compositional balance.
What is the limitation mentioned in the video regarding the use of negative prompts?
-The limitation mentioned is that mid-journey can be 'tricky' with negative prompts, sometimes not removing the specified elements as expected, such as when trying to remove a hat from an image that is present in the reference image.
What alternative method was used in the video to achieve the desired result when negative prompts were not effective?
-When negative prompts did not effectively remove the hat from the image, the video demonstrated the use of photobashing, where Clint Eastwood's face was added onto a body from a different image to achieve the desired result.
Outlines
🎨 Understanding Image and Text Weights in Mid-Journey Prompting
The video begins with an introduction to two powerful but often misunderstood techniques in mid-journey prompting: Image Weight and Text Weight. The speaker aims to clarify how these weights function and how they can be combined to control the output of generated images. The explanation includes a discussion on the limitations of weights and strategies to overcome them. The process of how prompts work in mid-journey is also briefly explained, detailing how keywords are identified and tokens are assigned to assemble the final image. The video illustrates the concept of text weights by showing how the importance of keywords can be adjusted using a colon colon notation followed by a number, which affects the composition of the generated image. The importance of formatting is emphasized, particularly the lack of space between the keyword and the colon colon, and the presence of a space after the weighted number.
📸 Experimenting with Image Weights and Reference Images
The second paragraph delves into the use of reference images in mid-journey, where an image URL is placed at the beginning of a prompt to influence the output. Image weights, denoted by a dash dash IW followed by a number between 0.5 and 2, are introduced to indicate the reliance on the reference image. The video demonstrates how varying the image weight can yield different results, showing examples where the style of an artist is incorporated into the generated image to varying degrees. The speaker also discusses the use of negative prompts to exclude certain elements from the final image, highlighting the challenges and workarounds when trying to remove specific features that the model is inclined to include.
🔍 Advanced Techniques and Photobashing for Desired Results
In the final paragraph, the speaker shares advanced techniques for maximizing image results and discusses the intricacies of negative prompting in mid-journey. Examples of attempting to remove an image element, such as a hat, are provided, demonstrating the model's tendency to include certain features despite negative prompts. The video concludes with a photobashing technique where the face of Clint Eastwood is combined with the body of another actor to achieve a desired look, showcasing an alternative method for obtaining specific results when direct text prompting is insufficient. The speaker encourages viewer engagement through likes and subscriptions and invites questions, comments, or suggestions.
Mindmap
Keywords
💡Midjourney
💡Text Weight
💡Image Weight
💡Prompt
💡Tokens
💡Compositional Balance
💡Negative Prompt
💡Photobashing
💡Reference Image
💡Illustrator Style
💡Dynamics
Highlights
Image Weight and text weight are powerful techniques in mid-journey prompting
Weights allow control over the compositional balance in generated images
Mid-journey assigns tokens to keywords in prompts, with more emphasis on earlier words
Text weights can be applied by using a colon colon followed by a number after a keyword
Image weights can be used with reference images to influence the style of the output
Weights can range from 0.5 to 2, indicating the reliance on the reference image
Negative prompts can be used to remove elements from the generated image
Photobashing can be used as an alternative method to achieve desired results when text prompts fail
The importance of correct formatting when applying text and image weights
Examples provided to illustrate the effects of different weights on the final image
The video demonstrates how to combine text and image weights for more control over output
Limitations of using weights and strategies to overcome them are discussed
The role of natural language in simplifying the process of achieving desired image results
The influence of the position of keywords in the prompt on the generated image
The concept of tokens and how they are distributed across the prompt
The use of decimal points in text weights for more precise control
The potential for experimentation with weights to achieve unique and creative results
The practical application of weights in a project about a fictional documentary