Best Character Consistency Midjourney Workflow!

Snowball AI
18 Mar 202407:41

TLDRIn this video, the creator at Snowball AI shares an exciting new feature from Midjourney, the Character Reference tool, which helps maintain character consistency across multiple images. The video demonstrates how to use this feature by creating a series of illustrations featuring the same character in various poses and angles. The process includes writing detailed prompts, using the character reference feature with a specific syntax (D-C followed by an image URL), and adjusting the character weight to focus on facial features or full character details. The video also addresses common issues like handling multiple characters in a scene and the need for post-processing in Photoshop to refine images. The creator emphasizes the power of combining character and style references to create a cohesive series of images, ideal for storytelling or children's books. The video concludes with tips on using Midjourney's zoom-out feature and the importance of lighting and Photoshop skills to finalize the artwork.

Takeaways

  • 🎉 The video introduces a new feature in Snowball AI called 'character reference' which helps in maintaining character consistency across multiple images.
  • 📈 The character reference feature works similarly to the 'style reference' feature, allowing users to input an image of a character and maintain its consistency.
  • ✍️ To use the character reference, one must include 'D-C' followed by the image URL of the character in the prompt.
  • 🔍 The character weight (CW) can be adjusted from 100 to 0 to control the level of detail in replicating the character's features.
  • 🚫 The feature is not designed for real people or photos and works best with characters created using Snowball AI.
  • 🧩 The character reference can be combined with the style reference feature to create a series of images with consistent style and character.
  • 🖼️ It's recommended to crop and edit character images to fix deformities before using them as references for better results.
  • 💡 Specific images of a character can be used for specific scenes to maintain the desired angle, pose, or expression.
  • 🖌️ Photoshop can be used to edit the background and make minor adjustments to the character for a more polished illustration.
  • 🔄 The zoom out feature in Snowball AI allows for adjusting the framing of the character from close-up to full body shots.
  • ✨ Lighting effects can add depth to the illustration and should be considered when creating character images.
  • 🔗 The video provides a link in the description for learning Photoshop, which is useful for beginners focused on similar use cases.

Q & A

  • What is the main focus of the video?

    -The main focus of the video is to demonstrate how to create consistent characters using the latest Midjourney feature, specifically the new character reference feature.

  • How does the character reference feature work?

    -The character reference feature works by allowing users to write 'D-C' followed by an image of their character to maintain consistency across multiple images. The reference strength can be modified using the '--CW' parameter, which stands for character weight, with a value between 100 and zero.

  • What is the default character weight if not specified?

    -The default character weight is 100 if it is not specified by the user.

  • What happens when the character weight is set to zero?

    -When the character weight is set to zero, the AI focuses only on the face, which is useful for changing outfits, hair, and adding accessories to the characters.

  • Is the character reference feature designed for real people or photos?

    -No, the character reference feature is not designed for real people or photos. It works best with characters made from Midjourney.

  • How can the character reference feature be combined with the style reference feature?

    -The character reference feature can be combined with the style reference feature by using both 'D-C' and '--SRE' followed by their respective image URLs to maintain both character consistency and a consistent style across images.

  • What is the recommended aspect ratio for creating a character design sheet?

    -The recommended aspect ratio for creating a character design sheet is 16:9, which provides a horizontal image with room for more outputs of the character.

  • Why is it important to crop each image separately and fix deformities before saving?

    -Cropping each image separately and fixing deformities before saving increases the chances of getting better results when using these images as references, especially if the character is the main focus and will appear in more than 10 illustrations of the story.

  • How can specific images of a character be used for specific scenes?

    -Specific images of a character can be used for specific scenes by selecting and using those images as references. This allows for the creation of images that match the desired angle, pose, or expression of the character for that particular scene.

  • What is the zoom out feature in Midjourney used for?

    -The zoom out feature in Midjourney is used to show more of the character's body in the generated image, which can be helpful when the reference images only show close-ups or medium close shots.

  • How does the video creator suggest using lighting in illustrations?

    -The video creator suggests using lighting by selecting an area and changing the blending mode to create a more realistic and visually appealing illustration. They also mention that lighting is often underrated but can significantly enhance the final result.

  • What additional tool or skill is recommended for editing and creating illustrations?

    -Photoshop is recommended for editing and creating illustrations. The video creator demonstrates how to use Photoshop for tasks such as removing backgrounds, adding missing elements, and adjusting the overall look of the character to fit the desired scene.

Outlines

00:00

🎨 Introducing Character Consistency with M Journey's New Feature

The video begins with the host expressing excitement about the latest M Journey feature, which is designed to create consistent characters across multiple images. The host discusses the challenges of character consistency in generative AI and how M Journey has been a top choice for generating images. The new character reference feature is introduced as a solution to maintain similar appearances of a character in different illustrations. The feature allows users to input a character image and modify the reference strength using the '--CW' parameter. The host also provides tips on using multiple images for references to capture different angles and details of the character. The video demonstrates how to use the feature by generating a character design sheet and then creating new images using the character sheet as a reference. The process of refining the generated images in Photoshop to achieve better consistency and style is also covered.

05:02

📚 Enhancing Character Images with Style and Character References

The host continues to explore the character reference feature by generating a new image of the previously created character in a different scenario, jumping over a fence. The video highlights the ability to combine character reference with style reference to maintain the artistic style across images. The host also addresses a common issue where multiple characters in a reference image can affect the outcome, but reassures that post-processing in Photoshop can resolve such issues. The video then moves on to creating an image of the character watching a movie and eating popcorn, demonstrating how to edit the background and adjust lighting to fit the scene. The host shares a trick about using the zoom out feature in M Journey to get full-body shots when the reference images are close-ups. Finally, the video concludes with a recommendation for a Photoshop tutorial for beginners, suitable for those looking to enhance their generative art skills.

Mindmap

Keywords

Character Consistency

Character consistency refers to the uniformity in the appearance and traits of a character across different illustrations or scenes. In the video, the speaker discusses the challenges of maintaining character consistency when using generative AI models, which often produce random results. The new character reference feature in Midjourney is presented as a solution to this issue, allowing for the creation of a character that appears consistently across multiple images.

Midjourney

Midjourney is an AI model for generating images. It is mentioned as the best AI model for generating images in the video. The speaker is excited about the new character reference feature in Midjourney, which is expected to enhance the process of creating illustrations with consistent characters.

Character Reference Feature

This is a new feature in Midjourney that allows users to input an image of a character and generate new images with the same character, maintaining consistency in appearance. The feature is used by writing 'D-C' followed by an image reference. It is highlighted as a significant advancement for creating illustrations with the same character in different poses and angles.

Style Reference Feature

The style reference feature is a tool in Midjourney that helps maintain a consistent style across various images. It was introduced a month prior to the character reference feature. By using an image as a reference, users can ensure that the style of the generated images matches the style of the reference image. In the video, this feature is combined with the character reference feature to create a cohesive series of images with both consistent style and character.

Character Weight (CW)

Character weight, denoted as CW, is a parameter in the character reference feature that allows users to modify the reference strength of a character image. It is set with a number between 100 and zero, with 100 being the default. A higher value focuses on replicating more details of the character, including face, hair, and clothes, while a lower value focuses primarily on the face, which is useful for changing outfits and accessories.

Generative AI

Generative AI refers to artificial intelligence systems that are capable of creating new content, such as images, music, or text. In the context of the video, generative AI is used to create illustrations and images of characters. The challenge with generative AI is its tendency to produce random results, which can make character consistency difficult to achieve.

Photoshop

Photoshop is a widely used software for image editing and manipulation. In the video, it is mentioned as a tool to further refine and edit the images generated by Midjourney. The speaker uses Photoshop to crop images, fix deformities, adjust backgrounds, and add details to the characters, which helps in creating more polished and consistent illustrations.

Children's Book Illustrations

Children's book illustrations refer to the visual representations used in children's books to convey the story or enhance the narrative. The speaker mentions creating illustrations in the style of children's book illustrations, which typically have a more cartoonish and 2D appearance. The character reference feature in Midjourney is used to generate a character design sheet for these illustrations.

Aspect Ratio

The aspect ratio is the proportional relationship between the width and the height of an image. In the video, the speaker mentions using an aspect ratio of 16:9, which provides a horizontal image with more room for multiple outputs of the character. This aspect ratio is chosen to accommodate the creation of a character design sheet.

Lighting

Lighting in the context of image editing and illustration refers to the way light interacts with the subject and the environment in an image. The speaker emphasizes the importance of lighting in creating a realistic and engaging illustration. In the video, lighting is adjusted to enhance the mood and depth of the character illustrations.

Zoom Out Feature

The zoom out feature in Midjourney allows users to expand the view of the generated image, revealing more of the character's body or surroundings. In the video, the speaker uses this feature to create images showing more of the character's body, which can then be edited and incorporated into wider scenes using Photoshop.

Highlights

Introduction to Snowball AI's process for creating consistent characters using the latest Midjourney feature.

The challenge of character consistency in generative AI due to the random nature of AI models.

The new character reference feature in Midjourney, which replicates face, hair, and clothing across images.

Adjusting character weight with the '--CW' parameter to focus on specific aspects of the character.

Tips for using character reference with non-photographic, Midjourney-created characters for best results.

Combining character reference with style reference for a cohesive series of images.

The importance of using multiple reference images to capture different angles of a character.

Creating a character design sheet by generating various poses and expressions of a character.

Cropping and editing individual images for better consistency when used as references.

Using specific images for specific scenes to maintain character consistency.

Generating a new image with the entire character sheet as a reference for consistency.

Combining character and style references to maintain both consistency and artistic style.

Addressing the issue of multiple characters in a reference image affecting the output.

Using Photoshop for editing backgrounds, adjusting character poses, and adding details.

The zoom out feature in Midjourney for capturing more of the character's body in the image.

Techniques for creating a consistent look across a children's book using Midjourney and Photoshop.

The impact of lighting on the final illustration and how to adjust it for a more realistic effect.

Recommendations for learning Photoshop quickly with a focus on use cases similar to the video's content.

Encouragement to subscribe for more content on character consistency and creative workflows.