Image Prompting Vs. Style Reference - Midjourney

Future Tech Pilot
5 Feb 202406:36

TLDRMidjourney has introduced a new feature called 'style reference,' which has led to inquiries about its distinction from 'image prompting.' The video explains this through three analogies: mathematical operations, cooking, and writing a story. Image prompting is likened to addition, where a reference image and words combine to create the final image. Style reference, on the other hand, is more akin to multiplication, where the style of the reference image is applied to the subject, resulting in a new creation that blends the two. The video uses examples such as a dog, a day at the beach, and a man wearing a red jacket to illustrate the concepts. The presenter, Nolan, aims to simplify AI learning and encourages viewers to like the video to help share the knowledge. The analogy of cooking is employed to further clarify the difference, with image prompting compared to making a sandwich and style reference to making a smoothie. The video concludes with a discussion on the 'D-ssw' parameter, which adjusts the influence of the reference image on the generation. The summary provides a clear understanding of when to use each option for creative purposes.

Takeaways

  • 📌 Image prompting is like addition, where the reference image and words in the prompt combine to create the final image.
  • 🔍 Style reference is akin to multiplication, where the image's style is applied to the subject, resulting in a new creation.
  • 🐶 Using a dog as an example, image prompting combines the image with the subject to form a new picture, whereas style reference applies the style to the subject.
  • 🏖️ For a 'day at the beach' prompt, image prompting overlays the image onto the prompt, whereas style reference blends the style of the reference image with the subject.
  • 🧀 Image prompting can be compared to making a sandwich, where the components (image and prompt) are clearly visible in the final product.
  • 🍹 Style reference is more like making a smoothie, where the essence of the components is blended together, not distinctly visible but still present.
  • 🧩 When using Lego bricks as an image prompt with a dragon, the final image is a clear combination of both, like a sandwich.
  • 🌂 If the goal is a specific style, like a yellow raincoat, image prompting will bring through the character of the image, while style reference will blend it more subtly.
  • ⛓️ For a medieval armor set, image prompting places the armor over the original image, whereas style reference will integrate the style of the armor into the subject.
  • 🦇 Writing a story analogy: image prompting focuses on the character (e.g., Batman), ensuring it appears in the results, while style reference is inspired by the vibe or genre, not specific characters.
  • 📈 The 'D-ssw' parameter allows for adjusting the influence of the reference image on the generation, with higher values resulting in images more closely aligned with the style of the reference.

Q & A

  • What is the main difference between image prompting and style reference as explained in the transcript?

    -Image prompting is likened to addition, where a reference image plus some words in the prompt equals the generated image. Style reference, on the other hand, is more like multiplication, where the image style is applied to the subject to create the generation, resulting in a blend of the two.

  • Can you provide an example of how image prompting works?

    -An example of image prompting is using a picture of a 'baby blue Memphis pattern' and adding a prompt for 'a dog'. The result is an image that combines the Memphis pattern background with a dog, reflecting the addition of elements.

  • How does style reference apply the style of an image to a subject?

    -Style reference takes the style of the reference image and applies it to the subject, creating a new image that reflects the style of the original image without necessarily including its specific elements. For instance, a 'day at the beach' reference image can be used to style a different subject, like a dog, resulting in an image that has the essence of a beach day.

  • What is the cooking analogy used to explain image prompting and style reference?

    -Image prompting is compared to making a sandwich, where the image and the prompt are combined directly (peanut butter plus jelly equals a PB&J sandwich). Style reference is likened to making a smoothie, where the essence and style of the image are blended with the subject, resulting in a more integrated outcome.

  • How does the writing a story analogy relate to image prompting and style reference?

    -In the context of writing a story, image prompting is akin to writing about a specific character, focusing on the details of that character. Style reference is more about capturing the vibe or genre of the story, which can inspire the overall feel of the narrative without focusing on specific characters.

  • What is the parameter D-ssw used for in style reference?

    -The parameter D-ssw, which stands for style weight, is used in style reference to determine the degree of influence the reference image has on the generation. A number between 0 and 1,000 can be assigned, with 1,000 resulting in a generation that is closely aligned with the style of the parent image.

  • How does the speaker, Nolan, aim to make learning AI straightforward?

    -Nolan aims to simplify the learning process of AI by providing clear explanations and analogies that relate complex concepts, such as image prompting and style reference, to everyday experiences like cooking or writing a story.

  • What is the purpose of using an image prompt with detailed words like 'a man wearing a red jacket and gold shoes'?

    -Using a detailed image prompt with specific words helps to generate an image that closely matches the description provided. However, the actual visual elements like the red jacket may not always be clearly depicted, but the overall concept is conveyed in the final image.

  • What is the significance of the 'style weight' parameter in the context of style reference?

    -The 'style weight' parameter allows for control over how much the style of the reference image impacts the final generation. It offers a way to fine-tune the output to achieve the desired balance between the subject and the stylistic influence.

  • How does the speaker use the concept of 'addition' to illustrate the process of image prompting?

    -The speaker uses the concept of 'addition' to illustrate how the elements of an image prompt and the subject come together to create a final image. It's a direct combination, where the elements are clearly identifiable in the resulting image.

  • What is the role of the reference image in style reference?

    -In style reference, the reference image provides the stylistic qualities that are applied to the subject. The specific content of the reference image becomes less important, and more emphasis is placed on transferring the style or the 'vibe' of the image to the subject.

  • How does the transcript differentiate between the visual outcomes of image prompting and style reference?

    -The transcript differentiates the visual outcomes by showing that image prompting results in a direct combination of the prompt and image, with elements clearly visible. In contrast, style reference produces an image that reflects the style of the reference image but blends it with the subject in a way that might not directly show the original elements.

Outlines

00:00

📈 Understanding Style Reference vs. Image Prompting

The video introduces a new feature called 'style reference' by Mid Journey and explains how it differs from 'image prompting.' The presenter uses three analogies to clarify the distinction: a mathematical addition, cooking, and writing a story. In the addition analogy, image prompting is likened to combining elements (image + prompt = result), whereas style reference is more like multiplication, blending the style of the image with the subject. The cooking analogy compares image prompting to making a sandwich, where ingredients are clearly visible, while style reference is akin to making a smoothie, where the ingredients blend into a unified style. Lastly, the story analogy describes image prompting as focusing on specific characters, whereas style reference is inspired by the mood or genre of the story. The presenter also mentions a parameter called 'D-ssw' that adjusts the influence of the reference image on the generation.

05:01

🎨 Applying Style Reference and Image Prompting Techniques

The second paragraph demonstrates the practical application of style reference and image prompting through examples. It shows how using an image of Batman as an image prompt results in images that clearly depict Batman, even when combined with the concept of a 'Ninja Turtle fight scene.' Conversely, applying style reference to the same Batman image results in a more abstract representation inspired by the vibe of the original image, such as a black and white manga style. The presenter emphasizes the flexibility of style reference and how it can be fine-tuned using the 'D-ssw' parameter, which allows for varying degrees of influence from the reference image, leading to a wide range of creative outputs.

Mindmap

Keywords

Image Prompting

Image prompting is a technique where a reference image is combined with descriptive words to generate a new image. It is likened to addition in the video, where the original image and the words in the prompt are added together to produce the final result. For example, using a picture of a 'Memphis pattern' and adding the word 'dog' results in an image of a dog with a Memphis pattern background.

Style Reference

Style reference is a method where the style of a reference image is applied to a subject to create a new image. It is compared to multiplication in the video, where the image's style is multiplied by the subject to generate the final image. Unlike image prompting, style reference does not replicate the entire picture but rather transfers the style, as seen when a 'day at the beach' reference image is used to style a different subject.

Addition

In the context of the video, addition is used metaphorically to describe the process of image prompting. It suggests a direct and literal combination of elements, where the original image and the prompt's descriptive words are clearly visible in the final image, maintaining their individual identities.

Multiplication

Multiplication serves as a metaphor for style reference in the video. It implies a blending of elements where the original image's style is applied to another subject, creating a new image that reflects the style of the reference image rather than its specific content. This is demonstrated when a 'dog' is styled with the pattern of a Memphis design.

Cooking Analogies

The video uses cooking analogies to explain the concepts of image prompting and style reference. A sandwich represents image prompting, where distinct elements (peanut butter and jelly) are combined to form a whole. Conversely, making a smoothie symbolizes style reference, where ingredients are blended to create a new essence that retains the style of the original components.

Writing a Story

The video compares image prompting to writing a specific character in a story, focusing on the character's attributes. For instance, using Batman as an image prompt results in images that directly feature Batman. In contrast, style reference is akin to being inspired by the genre or vibe of the story, which can lead to images that capture the essence of the story without featuring the original characters.

Nolan

Nolan is the presenter of the video, who aims to simplify the learning of AI concepts. His role is to explain the differences between image prompting and style reference using various analogies and examples, making the complex topic more accessible to viewers.

AI Generation

AI generation refers to the process of creating new images through artificial intelligence, using techniques like image prompting and style reference. The video demonstrates how different methods of input (image and words) are used by AI to generate unique outputs that can vary in style and content.

Parameter D-SSW

The parameter D-SSW, which stands for 'style weight,' is a tool used in style reference to control the influence of the reference image on the generated image. A value between 0 and 1000 determines the strength of the style transfer, with higher values resulting in a more pronounced style from the reference image.

LEGO Bricks

LEGO bricks are used as an example in the video to illustrate the concept of image prompting. When image prompted with a 'dragon,' the result is a dragon placed on top of LEGO bricks, maintaining the distinct elements of both the prompt and the original image.

Manga Inspired Visuals

Manga inspired visuals are a style outcome mentioned in the video when discussing the results of style reference. It refers to the aesthetic that is reminiscent of manga or comic book art, which can be achieved by applying the style of a manga image to a different subject.

Highlights

Midjourney introduces a new feature called 'style reference'.

Image prompting is compared to addition, where a reference image plus words equals the generated image.

Style reference is likened to multiplication, where the image's style is applied to the subject.

An example given is a 'baby blue Memphis pattern' used as an image prompt resulting in a specific generated image.

In style reference, the pattern isn't exact but maintains the same style, as demonstrated with a dog example.

Image prompting focuses on the details within the entire picture, while style reference sees the bigger picture and how it was made.

A cooking analogy is used to explain the difference: image prompting is like making a sandwich, and style reference is like making a smoothie.

The essence of the image is maintained in style reference, even when the original image elements are not directly visible.

A wide-angle product photo of a yellow raincoat is used to demonstrate how image prompting adds character to the generated image.

Style reference allows for a blend of the reference image's style with the subject, as shown with a medieval armor set example.

Writing a story analogy is used to further explain the difference, where image prompting focuses on the character, and style reference on the vibe or genre.

The parameter 'D-ssw' is introduced, which stands for style weight and influences how much the reference image affects the generation.

At a style weight of 1,000, the generated image is more closely aligned with the reference image's style.

The video provides clarity on when to use image prompting and style reference based on the desired outcome.

The presenter, Nolan, aims to make learning AI straightforward and encourages viewers to like the video for more content.

Additional information on how to expand prompts for better results is available in another video by the presenter.

The video concludes with a reminder to take care and a promise to see viewers in the next video.