How to Write the Most Accurate Midjourney Prompts - /describe Beginner Tutorial

Future Tech Pilot
9 Jun 202307:16

TLDRThe video script discusses the process of using the 'describe' feature in Discord's Mid-journey to generate prompts for images. It explains how to upload an image, select from generated prompts, and refine results by adding the original image as a reference, adjusting weights, and modifying stylize and chaos values for more creative freedom. The video also mentions a prompt pack available for purchase and highlights some quirks of the describe feature, such as its handling of made-up artist names and aspect ratios.

Takeaways

  • 🖼️ Utilize the 'describe' feature in Discord to generate prompts for an image saved on your computer by typing 'forward slash describe'.
  • 🔄 After using 'describe', you'll be presented with four different prompt options based on the image.
  • 📸 To refine the prompts, add the original image as a reference by uploading it and copying its address to include in the prompt.
  • 🔧 Adjust the importance of the reference image by adding a weight ('dash dash IW') with a number between 0.5 and 2 to control its influence on the generated image.
  • 🎨 Modify the 'stylize' or 'chaos' values to control the creativity and variety in the generated grid of images.
  • 🔄 Experiment with 're-roll' in the describe feature to get additional prompt variations.
  • 🌐 Be aware that the 'describe' feature uses a model not native to Mid-journey, which may interpret made-up names or descriptions.
  • 📊 Aspect ratios may not be exact due to Mid-journey output rounding to the nearest 32-pixel value.
  • 💡 The 'describe' feature can accurately describe an image but may not perfectly recreate it, especially if the image contains text.
  • 💼 For convenience and inspiration, consider using a prompt pack created by experienced users, which includes a variety of prompts and examples.

Q & A

  • What is the primary purpose of the 'describe' feature in Discord as mentioned in the transcript?

    -The 'describe' feature in Discord is used to generate descriptions or prompts for images saved on your computer. By typing 'forward slash describe', users can upload an image and receive different prompt options based on the image's content.

  • How can the reference picture be used more effectively in the 'describe' feature?

    -The reference picture can be used more effectively by adding it as an image prompt. This is done by uploading the picture directly into Discord, expanding it, copying the image address, and pasting it into the 'describe' prompt. This provides a foundational picture for the prompts, making the generated images closer to the original.

  • What is the significance of adjusting the weight (IW) of the reference picture in the 'describe' feature?

    -Adjusting the weight (IW) of the reference picture allows users to control the importance of the reference image in the generation process. A lower weight (e.g., 0.5) means the reference picture will have less influence, while a higher weight (e.g., 2) means the reference picture will be more significant than the textual prompt, resulting in images that more closely resemble the original.

  • What are the effects of changing the 'stylize' and 'chaos' values in the 'describe' feature?

    -The 'stylize' value affects how closely Mid-journey follows the prompt. A lower 'stylize' value (e.g., 0) means a more literal interpretation, while a higher value (e.g., 400) allows for more creative freedom. The 'chaos' value introduces variety into the generated grid. A 'chaos' value of 0 (default) results in similar-looking generations, while a higher value (e.g., 100) produces significantly different images.

  • How does the 're-roll' option in the 'describe' feature work?

    -The 're-roll' option allows users to generate a new set of four prompts based on the current input. These new prompts will be similar but not identical to the previous ones, offering more options for users to refine their desired output.

  • What is the significance of aspect ratios in the 'describe' feature?

    -Aspect ratios can affect the output of the 'describe' feature. However, Mid-journey outputs round to the nearest 32-pixel value, so using an aspect ratio like 16:9 may not result in the exact dimensions due to this rounding process.

  • How can made-up artist names be generated by the 'describe' feature?

    -The 'describe' feature has the capacity to interpret and generate prompts for made-up artist names. Even though the names do not correspond to real artists, the feature can still create descriptions based on these虚构 names.

  • Why might the 'describe' feature read text in an image but not recreate it in the generated output?

    -The 'describe' feature uses a separate model to interpret images, which is not inherently part of Mid-journey's generation capabilities. While it can understand and describe text within an image, it may not accurately recreate that text in the generated output due to the limitations of the generation model.

  • What is the purpose of the prompt pack mentioned in the transcript?

    -The prompt pack is a collection of 51 favorite prompts with 69 total example images created by the speaker. It is designed to save users time and provide them with a ready-to-use set of prompts for generating images with Mid-journey.

  • How can users benefit from using the 'describe' feature with a combination of reference pictures, weights, stylize and chaos values?

    -By using a combination of reference pictures, weights, stylize and chaos values, users can achieve a higher level of customization and control over the generated images. This allows for a more tailored output that closely matches the user's vision and desired style.

  • What is the importance of understanding the limitations of the 'describe' feature?

    -Understanding the limitations of the 'describe' feature is crucial for setting realistic expectations and troubleshooting when the generated images do not meet the desired outcome. It helps users to adjust their approach and better utilize the feature to achieve the best possible results.

Outlines

00:00

🖌️ Utilizing the Describe Feature in Discord for Image Prompts

This paragraph discusses the process of using the describe feature in Discord to generate prompts for an image. It explains that by having the image saved on your computer, you can utilize the /describe command to receive different prompt options. The user can select one of the provided options or upload the original image as a reference to get more accurate results. The paragraph also covers the use of weights to emphasize the importance of the reference image and adjusting stylize or chaos values to control the level of creativity and variety in the generated images.

05:01

🎨 Enhancing Image Prompts with Additional Techniques

The second paragraph delves into further techniques to refine image prompts. It mentions the possibility of using the re-roll feature for additional prompt variations and acknowledges the contributions of Squire Zed from Discord. The paragraph also humorously points out that sometimes the describe feature might mention non-existent artist names. It explains the aspect ratio discrepancies due to Mid-journey's rounding to the nearest 32-pixel value and concludes with a note on the limitations of the describe feature, particularly its inability to recreate text from an image, despite recognizing it.

Mindmap

Keywords

💡Describe feature

The 'describe feature' refers to a tool within the Mid-journey platform that can analyze and describe images. It is used to generate prompts based on an uploaded image. In the video, the describe feature is central to the process of creating art by interpreting and transforming the visual elements of an image into a prompt, which then serves as a foundation for generating new images. It is demonstrated how the describe feature offers different options for prompts based on the image's content, such as 'a pink and gold robotic person' or 'futuristic female portrait in pink and blue colors'.

💡Discord

Discord is a communication platform where the Mid-journey community interacts and utilizes various tools, including the describe feature. In the context of the video, Discord is the medium through which users can upload images and interact with the describe feature to generate art prompts. It serves as a social and creative hub for users to share their experiences and results.

💡Image prompt

An 'image prompt' is a visual input used to guide the generation of new images or art. In the video, the term is used to describe the original image that users want to use as a reference for creating new content. The image prompt is essential in the process of transforming the visual elements of the reference image into a textual description, which then serves as the basis for generating new images.

💡Weight

In the context of the video, 'weight' refers to the importance or influence given to the reference image in the image generation process. By adjusting the weight with a numerical value between 0.5 and 2, users can control how closely the generated images should resemble the reference image. A lower weight means the reference image has less influence, while a higher weight increases its significance.

💡Stylize value

The 'stylize value' is a parameter that controls the level of adherence to the prompt in the image generation process. A lower stylize value (e.g., zero) means the generated image will follow the prompt more closely and literally, while a higher value (e.g., 400) allows for more creative freedom and aesthetic enhancement, potentially resulting in images that are more visually appealing but less literal.

💡Chaos value

The 'chaos value' is a parameter that introduces variety and randomness into the image generation process. By adjusting the chaos value (from 0 to 100), users can control the level of diversity among the generated images. A chaos value of zero results in images that are similar, while a value of 100 creates images that are completely different from each other.

💡Re-roll

The term 're-roll' in the video refers to the action of generating new prompt options using the describe feature. It allows users to obtain additional variations of prompts based on the original image, providing more options to refine and hone in on the desired style or outcome.

💡Aspect ratios

Aspect ratios are the proportional relationships between the width and height of an image. In the video, it is mentioned that Mid-journey's outputs round to the nearest 32-pixel value, which means that the exact aspect ratio specified (e.g., 16 by 9) might not be achieved, resulting in slightly different proportions.

💡Mid-journey

Mid-journey is an AI-based platform for creating images and art through the use of prompts and reference images. It is the main subject of the video, where the speaker discusses various features and techniques for using the platform effectively. The term encompasses the tools and processes involved in generating new images based on user inputs and reference materials.

💡Prompt pack

A 'prompt pack' is a collection of pre-made prompts and example images created by experienced users of the Mid-journey platform. In the video, the speaker mentions creating their own prompt pack containing 51 favorite prompts with 69 total example images, which is available for purchase on their website. The purpose of a prompt pack is to save time for users and provide inspiration for creating new images.

Highlights

The process of writing a prompt for an image involves using the describe feature in Discord.

To use the describe feature, you need to have the image saved on your computer and then type 'forward slash describe' in Discord.

After using the describe feature, you are given four different options based on the image.

If the initial results do not closely resemble the original picture, you can add the original reference picture as an image prompt.

To add the reference picture, upload it to Discord and copy the image address to include in your describe prompts.

Adjusting the weight of the reference picture with 'dash dash IW' and a number can control its influence on the generated image.

The stylize and chaos values can be adjusted to control how closely Mid-journey follows the prompt or allows creative freedom.

Lowering the stylized value increases literalness, while raising it allows for more creative interpretations.

Adding chaos value introduces variety in the generated grid, with a maximum value of 100.

Even with a complex prompt and reference picture, the describe feature may not perfectly recreate the image, as it uses a separate model.

The describe feature can sometimes interpret made-up artist names, showing its adaptability.

Aspect ratios in the describe feature output may not match the input exactly due to rounding to the nearest 32-pixel value.

The describe feature offers a way to enhance image generation by providing a foundational picture and adjusting various settings.

The creator has made a prompt pack available for purchase, containing favorite prompts and examples to save time and guide users.

The re-roll function on the describe feature provides additional options for refining the style of the generated images.