Krita AI, Constant faces using Lora Models

26 Dec 202312:14

TLDRIn this informative video, the creator demonstrates how to achieve consistent facial imagery using Creer with AI diffusion and Lora models. The process involves setting up a new style or model, selecting a base model, and applying Lora files trained on personal artwork. The creator also discusses the importance of avoiding head inclusion in clothing Lora files to prevent facial override and suggests using names in prompts for character consistency. The video provides valuable insights for artists seeking to generate repetitive, high-quality character images.


  • ๐ŸŽจ Use Creer with AI diffusion plug-in to create consistent faces in images.
  • ๐Ÿ“ˆ For advanced users, train a model using your own artwork with Lora Models on Civit AI.
  • ๐Ÿ” Select a model and add Lora files to apply specific styles or faces to your images.
  • ๐Ÿ–ผ๏ธ Adjust the strength of the Lora influence for fine-tuning the output image.
  • ๐Ÿ“ Include a detailed description when adding a new style or model to ensure consistency.
  • ๐Ÿง‘โ€๐Ÿฆฑ To maintain character consistency, use the same character name and attributes in every prompt.
  • ๐Ÿ‘• When creating Lora models for clothing, ensure the head is not included to avoid face overriding.
  • ๐Ÿ”„ Generate multiple images to ensure consistency and use them for creating animations or other purposes.
  • ๐Ÿ“‰ Increase quality settings like 4K resolution and adjust the bookcase weight for better image results.
  • ๐Ÿ” Use the latest plug-ins to add weight to specific attributes and control their prominence in the image.
  • ๐Ÿš€ The presenter is disabled and has benefited from AI advancements, highlighting the positive impact of technology on individuals.

Q & A

  • What is the main topic of the video?

    -The main topic of the video is creating a consistent image of someone's face using Creer with AI diffusion and Lora models.

  • What is the recommended approach for beginners to start with Lora models?

    -Beginners are advised to first watch a video on how to create an imagery face using Queer, SS before attempting to apply Lora models to their images and models.

  • How does one set up a new style or model in Creer?

    -To set up a new style or model in Creer, click on the cogs, open the menu, click on the plus sign, and type in the name of the new style or model. Then, select the desired model to use.

  • What is the purpose of using Lora models in this process?

    -Using Lora models helps in applying consistent styles to images and models, which is crucial for maintaining the same character appearance across different images.

  • How does the video creator ensure consistency in the face across multiple renders?

    -The video creator ensures consistency by using a trained Lora model of their face, adjusting the strength to 90%, and applying it to the images.

  • What is the advice given for creating Lora models on clothing?

    -For clothing, it is recommended to train a Lora model without the head, making sure to crop the head off, to avoid the clothing model overriding the face.

  • How can one improve the quality of the generated images?

    -One can improve the quality by adding high-quality 4K resolution and adjusting the weight of certain elements using the latest plugins.

  • What is the video creator's personal experience with AI and their condition?

    -The video creator is disabled with a bugged arm and body. They have been able to continue their work thanks to AI, which has existed since at least August 2022.

  • What feature does the video creator wish to see in future AI improvements?

    -The video creator wishes to see the ability to do an alpha of a translucent background, allowing characters to be generated with no background, making layering easier.

  • How can one maintain consistency in AI-generated characters?

    -To maintain consistency, one can use the same character name in every prompt, ensuring the AI generates the same character with slight variations.

  • What is the importance of subscribing and liking the video according to the creator?

    -Subscribing and liking the video helps the YouTube algorithm to promote the content to more viewers, which in turn supports the creator and the community.



๐ŸŽจ Introducing AI Diffusion in Art Creation

The paragraph introduces the use of AI diffusion for creating a consistent image, specifically a face, using Creer with an AI plug-in. The speaker plans to demonstrate the setup process and provides a link for viewers to learn about creating imagery faces with another program called Queer. The speaker acknowledges that the Queer program might be challenging and recommends viewers to familiarize themselves with it before proceeding. The focus is on applying Lwas to images and models to achieve consistency, which is beneficial for those with advanced computing skills.


๐Ÿ–Œ๏ธ Step-by-Step Guide to Applying Styles in Creer

The speaker provides a detailed walkthrough on how to use the Creer software to apply styles to images. The process involves naming a new style, selecting a model, and importing Luras, which are AI-trained styles. The speaker emphasizes the importance of using Luras trained on specific elements like clothing without including the head to avoid overriding the main face model. Tips on achieving consistency in images, such as adjusting the strength of the Lura and using high-quality settings, are shared. The speaker also discusses the potential of using AI to generate characters for animations and the challenges of maintaining consistency across multiple renders.


๐Ÿš€ Improving Efficiency and Accessibility with AI

In this paragraph, the speaker reflects on the personal impact of AI on their creative process, given their disability. The speaker highlights the advancements in AI since August 2022 and expresses hope for future improvements, such as the ability to create translucent backgrounds for easier layering in image creation. The speaker also mentions a potential solution for background removal using AI add-ons and encourages viewers to watch a related video for more information on managing multiple Python versions to avoid software conflicts.



๐Ÿ’กConsistent image

The term 'consistent image' refers to a uniformly recognizable depiction of a subject, such as a person's face, across multiple digital representations. In the context of the video, it means creating a stable and repeatable visual representation of the artist's face using AI and Creer with AI diffusion plugin. The goal is to ensure that every generated image or render of the face remains identifiable and similar to the original, maintaining the same features and expressions throughout different artworks or designs.


Creer is a digital tool or software platform mentioned in the video that is used in conjunction with AI diffusion plugins to create and manipulate images. It is a tool that allows artists to apply styles and models to their images, thus generating new visuals that are consistent with the desired aesthetic. In the video, the artist uses Creer to apply Lora models to their face to achieve a consistent look in their artwork.

๐Ÿ’กAI diffusion

AI diffusion is a technique used in generative art and image processing that involves the use of artificial intelligence algorithms to create new images by diffusing or transforming existing ones. In the video, the artist uses an AI diffusion plugin with Creer to generate consistent facial images. This process typically involves training the AI on a dataset and then using the trained model to produce new images that reflect the learned patterns and styles.

๐Ÿ’กLora models

Lora models, or Lora files, are specific types of AI models that can be used to apply certain styles or characteristics to images. They are trained on particular datasets, such as an artist's own artwork, and can be used to ensure that generated images maintain a consistent look or style. In the video, the artist uses Lora models to apply their own facial features consistently across different images.

๐Ÿ’กRealistic Vision

Realistic Vision is mentioned as a model that the artist has selected from Civid AI to use in conjunction with their Lora model. It is likely a pre-trained AI model that helps in generating images with a realistic visual style. The use of this model contributes to the creation of lifelike and believable images in the artist's work.


In the context of the video, 'styles' refer to the visual characteristics or artistic expressions that are applied to images using AI models and plugins like Lora. These styles can dramatically change the overall look and feel of an image, making it unique or consistent with a particular artistic vision. The artist in the video is using styles to ensure that their face is represented in a consistent manner across multiple images.


In the context of the video, 'strength' likely refers to the intensity or degree to which an AI model or Lora file is applied to an image. Adjusting the strength can affect how prominently the applied style or model features in the final render. The artist in the video adjusts the strength to 90% to ensure that their face is prominently and consistently depicted in the generated images.


In the video, 'description' refers to the set of characteristics or attributes that the artist inputs to guide the AI in generating the desired image. This can include details about the subject's appearance, such as hair color, facial features, and clothing. The description helps the AI understand what specific elements to incorporate and prioritize in the creation process.

๐Ÿ’กCivid AI

Civid AI is mentioned as a source for pre-trained AI models, including Lora models, that can be used to apply styles to images. It is likely a platform or service that provides artists and users with a variety of models to work with, enabling them to create images with different styles and aesthetics.


In the context of the video, 'consistencies' refer to the uniformity and repetitive accuracy with which an AI generates images based on a given model or style. Achieving consistencies is important for artists who want their characters or subjects to be easily recognizable and maintain a specific look across various artworks or designs.


Animation, as discussed in the video, refers to the process of creating motion pictures or visual effects with a sequence of images. The artist mentions using AI and Creer not only for static images but also for animation, suggesting that the consistent facial representations created can be used in dynamic visual content.


The video demonstrates how to create a consistent image of a person's face using Creer with AI diffusion plug-in.

The process is suitable for advanced users with computing skills to create a model for generating faces.

The video provides a link to learn how to create an imagery face using a different program, Queer.

The user is guided through setting up a new style or model in Creer, starting with naming the style.

The model 'Realistic Vision' from Civa AI is used, and Lora files are added to customize the model.

The user's own trained model and face are incorporated into the Lora file for personalization.

Adjusting the strength of the Lora to 90% ensures a strong influence on the generated image.

The video explains how to apply Lora files to images and models for consistent results.

A description of the face is added to guide the AI in generating consistent facial features.

The video shows how to use the same technique for clothing by training a Lora on clothing from the neck down.

Consistent faces can be generated repeatedly by using the same Lora file and settings.

The video discusses the importance of not including the head in clothing Lora files to avoid facial feature overrides.

The user can generate multiple images with different positions and styles using the trained Lora file.

The video shows how to increase the quality of the generated images by adjusting settings like resolution and bookcase.

The AI tends to stick to the first generated image's features, which can be leveraged for consistency.

Naming characters and using them in every prompt helps maintain consistency across different renders.

The video touches on the potential for future improvements, such as the ability to generate images with translucent backgrounds.

The user shares their personal experience with using AI to overcome physical disabilities and limitations.

The video concludes with a call to action for viewers to like, subscribe, and turn on notifications for future content.