Early Access | Midjourney gets outperformed in character consistency!

The Reel Robot
14 May 202406:24

TLDRThe video provides an in-depth look at Leonardo AI's character consistency tool, which is considered one of the best for maintaining character consistency. The host demonstrates how to use the tool effectively by inputting a human face and adjusting settings such as photorealism and Alchemy mode. The character reference feature allows for customization of the character's aesthetic, including age, race, gender, and clothing, and can be used across different settings and outfits. The video also compares Leonardo AI's tool with Midjourney's, explaining the differences in how character references are used and how to adjust weights for character and style references to achieve desired outcomes. Additionally, the host touches on combining character and visual style references to create a unique aesthetic and offers tips for further enhancing AI video skills.

Takeaways

  • 🎨 **Leonardo AI's character reference tool** is considered the best for character consistency by the user.
  • 📷 **Photo Realistic (PR)** settings are optimal for the character reference feature, working best with human faces.
  • 🔍 **Alchemy mode** is used with a 16x9 image aspect ratio for the character reference tool.
  • 👔 **Character Reference Strength** can be set high to allow for simple prompts and generate character images with specific aesthetics.
  • 🎥 **Use Cases** include film, advertisement, music video, and storytelling, where characters can be placed in different outfits and settings.
  • 🔗 **Midjourney's Character Reference Tool** requires a different approach, using a prompt with a link to the character image.
  • 📈 **Character Weight (CW)** in Midjourney should be set below 100 to allow for variations like different outfits while maintaining the character's face.
  • 🌟 **Combining Style and Character References** can merge a character with a different visual style, creating a unique aesthetic.
  • 📚 **Pro Tip**: Using a fashion shot as a style reference can influence the background, clothing, and overall integration.
  • 🔄 **Style Weight (SW)** in Midjourney should be adjusted to blend the style with the character to the user's preference.
  • 🔧 **Image Weight (IW)** can be used in Midjourney to increase the influence of the character image in the prompt.

Q & A

  • What is the main purpose of Leonardo AI's character reference tool?

    -Leonardo AI's character reference tool is designed to maintain character consistency, especially with human faces, and to allow users to create images of characters in different outfits and settings based on a reference image.

  • How does the character reference feature work in Leonardo AI?

    -The character reference feature in Leonardo AI works by dragging an image of a face into the image guidance section, selecting 'character reference', setting the strength to high, and providing simple cues about age, race, gender, and clothing to generate consistent character images.

  • What settings are optimal for using Leonardo AI's character reference tool?

    -The optimal settings for Leonardo AI's character reference tool include having 'photo real B2' and an 'XL model', with 'Alchemy mode' turned on and a 16x9 image ratio.

  • How does the aesthetic dropdown in Leonardo AI affect the output image?

    -The aesthetic dropdown in Leonardo AI allows users to choose a style for the image. For example, selecting 'cinematic' will make the output look more like a movie scene, while 'portraits' will give a photography-like appearance, based on personal preference.

  • What is the significance of setting the strength to high when using character reference in Leonardo AI?

    -Setting the strength to high in Leonardo AI's character reference tool allows for a more consistent character output. Even with simple prompts regarding age, race, gender, and clothing, the AI can generate accurate and consistent character images.

  • How does Midjourney's character reference tool differ from Leonardo AI's?

    -Midjourney's character reference tool requires users to drag and drop an image and enter a prompt with 'Das CF' followed by a link to the image. It also has a character reference weight that defaults to 100, which can be adjusted to allow for different outfits while maintaining the character's face.

  • What is the character weight in Midjourney and how can it be adjusted?

    -The character weight in Midjourney is a setting that defaults to 100, which tries to match the outfits and other elements in the reference image almost exactly. Users can adjust this by entering 'CW' followed by a number lower than 100 to allow for variations while keeping the character's face consistent.

  • How can users combine a character reference with a style reference in Leonardo AI?

    -In Leonardo AI, users can combine a character reference with a style reference by dropping a style image into the second image slot, setting the type to 'style reference'. This allows the AI to maintain the same character while incorporating the style traits from the other image.

  • What is the purpose of adjusting the style weight in Midjourney when combining character and style references?

    -Adjusting the style weight in Midjourney with 's sw' followed by a number helps to control the blend between the character and style references. It allows users to fine-tune the influence of the style on the character image until it reaches a desired aesthetic.

  • What is the recommended approach when using both character and style references in Midjourney?

    -When using both character and style references in Midjourney, it's recommended to remove the character reference and place the link to the character image at the start of the prompt, followed by 'IW' for image weight and a number between 1.5 and 2 to balance the character and style elements effectively.

Outlines

00:00

🤖 Introduction to Leonardo AI's Character Reference Tool

The speaker introduces Leonardo AI's character reference tool, which they consider to be an exceptional resource for maintaining character consistency. They offer a tutorial on how to effectively utilize this new feature. The tool is particularly effective with human faces and can be adjusted with various settings to achieve optimal results. The speaker demonstrates how to use the tool by uploading a face image and selecting character reference with high strength. They also explain how to use simple cues to guide the AI in generating images with specific age, race, gender, and clothing descriptions. The tool's utility is highlighted for applications in film, advertisements, music videos, and storytelling, where it can reliably place characters in different outfits and settings with desired facial expressions.

05:00

🎨 Comparing Leonardo AI with Mid Journey for Character Consistency

The speaker compares the character reference tool of Leonardo AI with Mid Journey, noting the differences in their approaches. With Mid Journey, users drag and drop an image and use a specific prompt format including 'Das CF' followed by a link to the image. The speaker points out that Mid Journey's character reference weight is set to 100 by default, which can cause the AI to focus too heavily on matching the image, potentially ignoring other aspects of the prompt. To counter this, they suggest adjusting the character weight ('--CW') to a number lower than 100, allowing for more flexibility in generating images with different outfits while maintaining the character's face. The paragraph concludes with a brief mention of improving AI video skills and learning from 'The Whiz' for those interested in creating AI movies.

Mindmap

Keywords

💡Early Access

Early Access refers to a privilege granted to a select group of users to use a product or service before it becomes available to the general public. In the context of the video, the narrator has been given special access to a new tool developed by Leonardo AI for maintaining character consistency in AI-generated images. This is significant because it allows the narrator to explore and demonstrate the tool's capabilities ahead of its official release.

💡Character Consistency

Character consistency is a concept in AI image generation where the same character or subject is depicted with a consistent appearance across different images or scenes. This is crucial for storytelling and maintaining audience engagement. The video's theme revolves around showcasing how Leonardo AI's new tool helps achieve this consistency by using character reference images to guide the AI in generating images.

💡Leonardo AI

Leonardo AI is the name of the artificial intelligence tool being discussed in the video. It is highlighted as a superior tool for character consistency in AI-generated images. The video script provides a tutorial on how to use this tool effectively, emphasizing its features and capabilities in maintaining a consistent character appearance across various outputs.

💡Character Reference Tool

A character reference tool is a feature within AI image generation software that allows users to input a reference image to guide the AI in creating consistent character depictions. The video demonstrates how to utilize Leonardo AI's character reference tool to ensure that the generated characters maintain a consistent look, which is essential for projects like films, advertisements, and music videos.

💡Photo Real

Photo Real refers to the setting in AI image generation that aims to create images that are incredibly realistic, resembling actual photographs. In the script, the narrator mentions having the 'Photo Real B2' setting enabled, indicating a preference for highly realistic image outputs when using Leonardo AI's character reference feature.

💡Alchemy Mode

Alchemy Mode is a term used within the context of the video to describe a specific setting or mode within Leonardo AI that enhances the AI's ability to generate images. The narrator mentions having Alchemy mode on, suggesting it as an optimal setting for using the character reference tool to achieve the best results.

💡Image Guidance

Image Guidance is a feature that allows users to guide the AI's image generation process by providing it with a reference image. In the video, the narrator explains how to use image guidance in Leonardo AI by dragging in a face image to base the character on, which helps the AI to generate consistent character images.

💡Aesthetic

In the context of AI image generation, an aesthetic refers to the visual style or look that the generated images will have. The video script mentions an aesthetic dropdown, where users can choose different styles like 'cinematic' or 'portraits' to influence the look of the generated images, adding a layer of customization to the character consistency.

💡Mid Journey

Mid Journey is another AI tool mentioned in the video for comparison with Leonardo AI. The script outlines how to use Mid Journey's character reference tool, highlighting the differences in approach and settings between the two tools. The comparison serves to illustrate the strengths of Leonardo AI's character consistency capabilities.

💡Character Weight

Character weight is a parameter in AI image generation that determines the influence of the character reference on the final image. The video explains that in Mid Journey, the character weight defaults to 100, which can cause the AI to focus too much on matching the reference image. The narrator advises adjusting the character weight to less than 100 to allow for variations like different outfits while maintaining the character's face.

💡Style Reference

Style reference is a concept used to introduce a specific visual style or aesthetic into the AI-generated images, in addition to the character reference. The video demonstrates how to combine a style reference with a character reference in Leonardo AI, allowing the character to adopt the style traits of another image, such as glowing eyes and facial markings, while retaining their identity.

Highlights

Leonardo AI's character reference tool is highly effective for maintaining character consistency.

Early access to Leonardo AI's tool is granted, showcasing its superiority in character consistency.

The tool works optimally with human face inputs and specific settings.

Photorealistic B2 and an XL model settings are recommended for best results.

Image guidance allows for the selection of character reference with high strength settings.

Aesthetic choices such as 'cinematic' or 'portraits' can be selected for different visual styles.

Simple character reference cues can be used for age, race, gender, and clothing descriptions.

The tool enables reliable character placement in different outfits and settings.

Facial expressions can be described for scenes to be reflected in the output.

Comparison with Mid Journey's character reference tool is made using Discord.

Mid Journey's character reference weight defaults at 100, which may ignore parts of the prompt.

Adjusting character weight (CW) to below 100 allows for different outfits while maintaining the face.

Combining character and visual style references can merge aesthetics with character traits.

Leonardo AI allows for style traits to be added to a character with adjustable strength settings.

Mid Journey also supports style references with character references, using specific parameters.

Style weight (-ssw) and image weight (--IW) can be adjusted in Mid Journey for the desired blend.

Leonardo AI and Mid Journey are compared for their effectiveness in character consistency.