Midjourneyで同じキャラクターを作成する方法！同じキャラを出しながら、ポーズや表情を変える！

In this section, the speaker introduces the challenge of creating consistent characters using MidJourney, a popular AI illustration tool. Despite its versatility in generating diverse illustrations, it struggles with producing the same character repeatedly. Even with identical prompts, the output varies significantly, making it hard to maintain the same character across different illustrations. The speaker promises to explain a method to ensure consistency, beginning with an example of a boy character. After generating an initial image, they show how running the same prompt yields different results, highlighting the difficulty. The speaker emphasizes the importance of consistency for use cases like creating a storybook where the protagonist’s appearance must remain the same.

This part explains the need to tweak the prompts in MidJourney to achieve consistent characters. The speaker introduces two key phrases: 'Multiple Poses & Expressions' to instruct the AI to generate characters with various poses and expressions, and 'No Outline' to ensure the illustrations have no borders. By using these phrases, the resulting images display a consistent character across different expressions and poses. The speaker then selects one of the images, refines it by enhancing its resolution using an external AI tool called 'Upscale Media,' and explains the process of saving and editing the image further in tools like Photoshop or Canva.

In this section, the speaker explains how to further work with the generated images. After editing and cutting out the individual character illustrations, the speaker ends up with several different versions of the character. The edited images are then uploaded back into MidJourney, where links to the images are copied and embedded into a new prompt. These image links, combined with the original prompt (but with the 'Multiple Poses & Expressions' phrase removed), help the AI reference the uploaded images to create new illustrations of the same character with slight variations.

Here, the speaker provides an example of how to use the newly generated prompt to make the consistent character perform specific actions. For instance, they instruct the AI to depict the boy character running in a side-view position by adding the phrase 'Running, Side.' The AI produces several running illustrations, and the speaker chooses the most suitable one. They emphasize how consistency is maintained through this method, with the same boy character appearing in various poses while preserving his design.

In this section, the speaker demonstrates how to make the character smile and face forward by modifying the prompt. They add 'Smiling Happily' and 'Front View' to the existing prompt, instructing the AI to create a smiling, front-facing version of the character. While some attempts may not be perfect, the speaker reassures the audience that repeating the process will eventually yield the desired result. This process ensures that the same character is consistently generated across different facial expressions and orientations.

The speaker concludes by summarizing the entire process. First, they generate multiple character illustrations using 'Multiple Poses & Expressions' and 'No Outline' prompts. Then, they increase the resolution of the images and edit them before uploading them back to MidJourney. After uploading, the links to the edited images are included in a prompt along with the original description, minus the 'Multiple Poses & Expressions' phrase. This allows the AI to reference the edited illustrations and consistently generate similar characters. Lastly, additional prompts like 'Running' or 'Smiling' can be added to make the character perform specific actions, ensuring consistency throughout.

The video wraps up with the speaker thanking the audience for watching and expressing hope that the tutorial will help them create consistent characters using MidJourney. The speaker highlights how this method solves one of MidJourney's key weaknesses, which is the inability to generate the same character across different illustrations, and encourages viewers to try the process themselves.