【🦜Midjourney以图生图】最新超详细新手入门教程！独家技巧公开！教你保持人物统一连续，用ChatGPT+AI绘画软件制作高质量故事短片赚取收入，怎么写提示词，垫图与权重参数—第4集｜暗夜飞行
TLDRThe video script introduces a method for generating consistent character images across different scenes using AI tools like ChatGPT and Midjourney. It emphasizes the importance of creating a unified character portrayal by using '垫图' (base images) and adjusting settings like image weight (iw). The tutorial walks through the process of generating a fairy tale story with a fox protagonist,细分故事 into scenes, and then using a formulaic approach to instruct Midjourney to produce images that maintain stylistic and character consistency. The script also highlights the need for iteration and patience to achieve the desired results.
Takeaways
- 📖 Utilize ChatGPT to generate a story plot, specifying the main character as a fox with certain limitations on secondary characters and settings.
- 🖌️ Break down the story into smaller segments, each corresponding to a scene that will be visualized as an image.
- 🎨 Use Midjourney for AI image generation, focusing on maintaining the continuity and consistency of the main character across different images.
- 🖼️ Prepare by generating images of the story's protagonist and a desired scene to be used as '垫图' (base images) for the AI to reference.
- 📝 Construct a detailed prompt for Midjourney, including the base images, scene setting, character name, key features, activity, and desired artistic style.
- 🔗 Understand the use of image weight (iw) in prompts to control the influence of the base images on the final result, with higher values leading to closer resemblances.
- 🔄 Use double colons (::) in prompts to separate different elements, allowing Midjourney to distinguish and prioritize them correctly.
- 📈 Iterate and refine the AI's output by selecting the most satisfactory images and further tailoring them to meet the story's requirements.
- 🎭 Choose an artist's style to unify the visual effects of the generated images, such as anime or specific artists like宫崎骏 (Hayao Miyazaki) or 新海诚 (Makoto Shinkai).
- 📚 The process involves a lot of trial and error, requiring patience and multiple iterations to achieve the desired outcome.
- 👥 Encourage viewers to try the method, and for those without access to Midjourney or GPT-4, suggest looking into shared hosting platforms.
Q & A
What is the main topic of the video script?
-The main topic of the video script is about using AI tools like ChatGPT and Midjourney to generate a consistent character across different images for storytelling, specifically in creating a fairy tale with a fox as the protagonist.
How does the video script suggest generating a story?
-The script suggests using ChatGPT to generate a story by providing it with specific prompts, such as asking for a short fairy tale with a fox as the main character and certain scene limitations.
What is the significance of 'padding' in the context of Midjourney?
-In the context of Midjourney, 'padding' refers to the process of using existing images as a reference or starting point for the AI to generate new images. This helps in maintaining the consistency of characters and scenes across different images.
What is the role of 'image weight' (iw) in Midjourney's image generation process?
-The 'image weight' (iw) in Midjourney's image generation process controls the influence of the padded image on the final result. A higher iw value means the final image will be more similar to the padded image, making it crucial for maintaining character consistency.
How does the video script explain the use of double colons in Midjourney's prompts?
-The double colons in Midjourney's prompts are used as a separator to distinguish different elements within the prompt. They help the AI understand that separate words or phrases should be treated as independent concepts rather than a single entity.
What is the purpose of mentioning artists in the Midjourney prompt?
-Mentioning artists in the Midjourney prompt helps to set the desired artistic style for the generated image. By referencing specific artists, the user can guide the AI to produce images in a particular aesthetic or style that aligns with the artist's work.
How does the video script address the challenge of achieving the desired results with AI-generated images?
-The script acknowledges that achieving desired results with AI-generated images may require iteration. It suggests using the AI's output as a starting point and refining the prompts through a process of trial and error to gradually align the generated images with the user's vision.
What is the importance of maintaining a consistent character across different images in storytelling?
-Maintaining a consistent character across different images is crucial in storytelling as it helps the audience recognize and connect with the characters throughout the narrative. It provides a sense of continuity and familiarity, enhancing the overall storytelling experience.
How does the video script suggest iterating the AI's output?
-The script suggests iterating the AI's output by selecting the most promising results and using them as a new starting point for further generation. This process involves refining the prompts based on the AI's previous output, adjusting settings, and adding more specific instructions to achieve the desired outcome.
What is the role of 'aspect ratio' in image generation?
-The 'aspect ratio' in image generation determines the proportion of the image's width to its height. It is an important parameter as it sets the canvas size and shape, influencing the composition and layout of the generated image.
What is the significance of the artist Simon Birch in the script?
-Simon Birch is mentioned in the script as an example of an artist whose style the user wants to emulate in the AI-generated images. The user aims for a fairy tale look with a手绘风格 (hand-drawn style), and referencing Simon Birch helps guide the AI towards this visual aesthetic.
Outlines
🎨 Advanced AI Art Creation Techniques
The narrator introduces a popular trend of combining ChatGPT's storytelling with AI art generation tools like Midjourney or DALL-E to create novel animations and images. They discuss the common challenge of maintaining consistency in character appearance across multiple images and propose a sophisticated approach endorsed by Midjourney officials. This method involves using 'foundation images' and adjusting image weights to ensure continuity of character roles across different scenes. The narrator outlines the process starting with generating a story plot using ChatGPT, breaking it down into scenes, and then creating images that correspond to these scenes while ensuring character consistency using specific techniques like image padding and weight adjustment.
🔍 Understanding Image Weight (IW) and Multi-Prompts
The narrator delves into the concepts of image weight (IW) and the use of multi-prompts to fine-tune AI-generated art. IW is explained as a parameter that determines the influence of a base image on the final artwork, with a range from 0 to 2, where a higher value means greater similarity to the base image. The narrator illustrates this with an example where adjusting IW alters the balance between elements of a cake and flowers. Additionally, the concept of multi-prompts, indicated by double colons (::), is introduced as a way to help Midjourney distinguish between different elements within a prompt, significantly affecting the outcome of the generated images.
🌟 Perfecting Character Consistency Across Scenes
The final paragraph covers the narrator's personal journey of applying the discussed techniques to create a cohesive series of images for a fairy tale story featuring a fox named Mark. They share insights on preparing base images, describing scenes succinctly, and selecting artistic styles to maintain stylistic unity across images. The process of iterative refinement through multiple image generations is emphasized as crucial for achieving desired outcomes. The narrator also mentions the practical aspects of accessing Midjourney and ChatGPT, suggesting shared platforms for users without individual accounts, and concludes by inviting viewers to appreciate the completed story, underscoring the importance of support and engagement from the audience for content creation.
Mindmap
Keywords
💡ChatGPT
💡Midjourney
💡DAll-E
💡Seed values
💡垫图 (Padding Images)
💡权重 (Weight)
💡Multi-prompts
💡艺术家 (Artist Style)
💡迭代 (Iteration)
💡故事大王 (Master of Stories)
💡角色连续性 (Character Consistency)
Highlights
The use of ChatGPT and AI drawing software like Midjourney or DAll-E is currently very popular for generating novel and animation plots and images.
Despite the simplicity of the process, maintaining character consistency in generated images is challenging.
The video introduces a method from Midjourney's official guide to maintain character uniformity across different images.
The first step involves using ChatGPT to generate a story plot, with the example being a fairy tale featuring a fox as the protagonist.
The story is then broken down into smaller segments, each corresponding to a scene for image generation.
Midjourney is used to generate images for different scenes while ensuring the main character's continuity and uniformity.
The importance of preparing images of the story's protagonist and scenes for use in prompts is emphasized.
The prompt formula involves using an image address for padding (垫图), setting the scene, describing the protagonist, and specifying the activity and artist style.
The 'iw' parameter controls the influence of the padded image on the final result, with higher values leading to closer resemblance.
Double colons in the prompt are used to separate different elements, allowing Midjourney to better distinguish between them.
An example is provided to illustrate how the 'iw' value affects the balance between the influence of the padded image and the textual description.
The video demonstrates the practical application of the formula by generating images for a fairy tale story with a red fox named Mark.
The process involves iterative refinement, as AI-generated images may not meet all desired effects in a single attempt.
The video creator shares their experience of selecting and refining over 100 images to find the most satisfactory ones for their fairy tale.
The video concludes with an invitation for viewers to try the method themselves and provides a link for those without access to Midjourney and GPT-4.
The video creator encourages viewers to appreciate the complete version of the fairy tale with their voiceover and to support their channel for more quality content.
