Midjourney v5 - Style Prompt Tips and Reference Tricks

Theoretically Media
21 Mar 202311:57

TLDRIn this tutorial, the host delves into the art of effective prompting for Midjourney V5, focusing on achieving desired images through specific instructions. They address common issues like ignored instructions and the challenge of getting full-body images, often resulting in waist-up shots due to the program's cinematic tendencies. The host shares a methodical approach to refining prompts using a formula and adjusting the prompt's structure to better guide the AI's output. They also experiment with different styles, including those of Frank Miller, to illustrate how to steer the AI towards particular artistic outcomes. The video serves as a practical guide for users looking to enhance their control over AI-generated images.

Takeaways

  • 😀 The video discusses techniques for improving image output from Midjourney V5 using style prompts and references.
  • 🔍 Eric Schlitzbeyer's comment about Midjourney ignoring certain instructions inspired the video.
  • 🎨 The speaker attempts to generate a specific image of a 10-year-old Viking girl based on Eric's detailed prompt.
  • 📐 The video highlights the challenges with getting full-body images due to Midjourney's cinematic composition tendencies.
  • 🤔 It's suggested that Midjourney V5 is more linguistic but still requires a mix of programming and literary language in prompts.
  • 👉 The speaker uses a prompt formula to adjust the image output, emphasizing the importance of prompt structure.
  • 🌟 By adjusting the prompt and emphasizing 'Style by Frank Miller,' the images start to align more with the desired style.
  • 👣 The video points out that Midjourney might ignore certain details like 'barefoot' if they're not commonly found in its training data.
  • 🖼️ Image prompting and mixing tools like Leonardo are recommended to achieve better results.
  • 🎬 The video shows a creative process of turning a photo into an illustration and then into a cinematic still.
  • 🎭 A photobash technique is suggested for getting Midjourney to produce images with specific emotions.
  • 📈 The importance of experimenting with different styles and artists is emphasized to achieve the desired outcome.

Q & A

  • What is the main topic of the video?

    -The main topic of the video is discussing tips and tricks for effective prompting in mid-journey V5, specifically focusing on posing and image references to control the output and achieve desired images.

  • Who inspired the creation of this video?

    -The video was inspired by a comment left by Eric Schlitzbeyer on the creator's last video, expressing frustration with mid-journey's ability to follow instructions and style prompts.

  • What is the issue Eric had with mid-journey?

    -Eric's issue with mid-journey was that many instructions were ignored, such as requests for full body pictures instead of portraits from the hip, and the style prompt being completely disregarded.

  • What is the general advice given for prompting in mid-journey V5?

    -The advice given is to speak naturally to mid-journey, although the creator suggests that it still requires a more programming-like approach rather than a literary one.

  • What is the significance of the 'style by' prompt in mid-journey?

    -The 'style by' prompt is significant because it is meant to influence the visual style of the generated image, such as emulating the style of specific artists like Frank Miller.

  • How does the aspect ratio affect the type of images generated by mid-journey?

    -The aspect ratio can influence the composition of the images. A 16-9 aspect ratio tends to result in more cinematic compositions, often showing waist-up or close-up shots rather than full body shots.

  • What is the prompt formula mentioned in the video?

    -The prompt formula mentioned is: /imagine cinematic still film by scene subject action set a link or shot, which can be adapted to achieve specific results.

  • What is the problem with the initial image generated from Eric's prompt?

    -The initial image generated from Eric's prompt did not meet the requirements of a full body shot and did not capture the Frank Miller style as requested.

  • How can the prompt be adjusted to emphasize a particular aspect, such as the style?

    -The prompt can be adjusted by placing more emphasis on a particular word using a colon colon (::) notation, which can increase the weight of that aspect in the generated image.

  • What is the role of image prompting in the process?

    -Image prompting involves uploading a reference image to mid-journey and using it to guide the style and composition of the generated image. However, it can lock the aspect ratio to that of the reference image.

  • What is the solution to changing the aspect ratio when using image prompts?

    -The solution is to use the canvas feature in Leonardo, where you can take the generated image and expand it into the desired aspect ratio by painting out sections of the image.

  • What is the creator's final suggestion for achieving specific emotions in generated images?

    -The creator suggests using a 'really bad photobash' in Photoshop to create a rough image that can be used as an image reference in the prompt to guide the emotion of the generated image.

Outlines

00:00

🖌️ Enhancing Image Prompting in Mid-Journey V5

The speaker discusses the challenges of getting specific images from Mid-Journey V5 and aims to provide tips for better control over output. Inspired by a comment from Eric Schlitzbeyer, the video addresses issues such as the AI ignoring full body image requests and style preferences. The speaker shares initial output results from Eric's prompt and notes the discrepancies, such as the lack of full body shots and the absence of Frank Miller's distinct style. They suggest that the AI's natural language leans towards cinematic compositions, which often favor waist-up or close-up shots. To correct these issues, the speaker proposes adjusting the prompt's logic and using a formula to guide the AI more effectively.

05:01

🔍 Refining Prompts for Style and Composition

Continuing the exploration of Mid-Journey V5's prompting capabilities, the speaker experiments with different strategies to achieve the desired image output. They use a formula that includes elements like 'cinematic still,' 'film by,' 'scene,' 'subject,' 'action,' 'set,' and 'shot' to guide the AI. By emphasizing 'full body' and 'ultra long shot' alongside an aspect ratio of 16:9, the speaker manages to get closer to the intended image. However, they note that certain details, like shoeless Vikings, are challenging due to the AI's training data. The speaker also tries using image prompting and other tools like Leonardo to adjust aspect ratios and experiment with different styles, including Ridley Scott's, to achieve varied and interesting results.

10:03

🎨 Combining Techniques for Character Emotion and Aspect Ratio

The speaker focuses on the difficulty of conveying specific emotions and aspect ratios in Mid-Journey V5's outputs. They demonstrate a technique involving a 'photo bash' in Photoshop to add character emotion, which when used as an image reference, yields more emotive results. The speaker also addresses the challenge of maintaining the desired 16:9 aspect ratio and shares a solution using Leonardo's canvas feature. This feature allows them to expand the image while maintaining the original style and elements. The video concludes with a call to action for viewers to like, subscribe, and share their thoughts in the comments, and the speaker, Tim, thanks the viewers for watching.

Mindmap

Keywords

💡Midjourney V5

Midjourney V5 refers to the fifth version of a software or tool, presumably used for generating images based on textual prompts. In the context of the video, it is the platform that the speaker is discussing and providing tips for. The video aims to help users control the output of Midjourney V5 to achieve desired images, as illustrated by the discussion on posing and image references.

💡Prompting

Prompting, in the context of this video, is the act of providing textual instructions to the Midjourney V5 software to generate specific images. The script discusses how to effectively prompt Midjourney V5 to get the desired results, such as full body shots or images in a particular style, which is central to the video's theme of controlling image output.

💡Eric Schlitzbeyer

Eric Schlitzbeyer is mentioned as the person who inspired the video through a comment on a previous video. His comment highlighted issues with Midjourney's ability to follow instructions, such as ignoring requests for full body pictures or specific styles. This individual's feedback serves as a real-world example of the challenges discussed in the video.

💡Full Body Shot

A full body shot refers to an image that captures a subject from head to toe. The script mentions that the prompt called for a full body shot of a Viking girl, but the initial output from Midjourney V5 was a waist-up shot. This discrepancy is used to illustrate the common issue of aspect ratio and composition in image generation.

💡Aspect Ratio

Aspect ratio is the proportional relationship between the width and height of an image or screen, typically given as two numbers separated by a colon. In the video, the aspect ratio is discussed in the context of image composition, where the 16:9 aspect ratio tends to produce more cinematic, waist-up shots rather than full body images.

💡Frank Miller

Frank Miller is a renowned comic book illustrator and writer known for his work on 'Sin City,' '300,' and 'The Dark Knight Returns.' The script uses his name as an example of a style that was requested but not achieved in the initial image outputs from Midjourney V5. This highlights the challenge of applying specific artistic styles to image generation.

💡Cinematic

Cinematic refers to the style or quality of cinema, often characterized by a visual composition similar to that of a movie. The video discusses how Midjourney V5 tends to produce cinematic compositions, which can affect the type of images generated, such as the prevalence of waist-up or close-up shots over full body shots.

💡Image Prompt

An image prompt is a reference image used to guide the image generation process. The script describes using an image prompt of a Viking warrior to influence the output of Midjourney V5. However, it also notes the limitation that the aspect ratio of the reference image can restrict the aspect ratio of the generated image.

💡Leonardo

Leonardo is another tool mentioned in the video, used for image manipulation and generation. The speaker uses Leonardo to adjust the aspect ratio of an image generated by Midjourney V5, demonstrating the use of multiple tools to achieve the desired image output.

💡Ridley Scott

Ridley Scott is a famous film director known for his work on period pieces and cinematic visuals. In the script, his name is used as a prompt in Midjourney V5 to generate images with a cinematic style inspired by his work, showing how specific keywords can influence the style of image generation.

Highlights

Midjourney V5's new prompting capabilities are discussed, focusing on posing and image references.

Eric Schlitzbeyer's comment inspires the video, highlighting issues with specific image outputs.

The video aims to provide tips for controlling output to achieve desired images.

Midjourney V5's prompt system is suggested to be more linguistic but still requires a programming approach.

A prompt formula is introduced to help achieve closer results to desired images.

The importance of the order in the prompt is emphasized for better results.

An example prompt is used to demonstrate the process of refining images in Midjourney V5.

The video shows attempts to correct the aspect ratio and style issues in generated images.

Frank Miller's style is attempted to be applied to a Viking girl image with mixed results.

The use of Leonardo and its canvas feature to adjust aspect ratios is demonstrated.

Experimentation with different styles and artists is shown to achieve better results.

The concept of 'frankness' in prompts and its effect on image generation is discussed.

A technique for adding specific emotions to characters using photo bashing is suggested.

The video concludes with a call to action for viewers to like, subscribe, and comment.

A method for combining tools and techniques is promoted for achieving desired image outputs.

The video provides a detailed walkthrough of the process of generating and refining images in Midjourney V5.