真摄影级出图!Midjourney V6 alpha测试视频 MJV6对比DALLE 3 谁优谁劣?使用Style raw参数在MJ中画出照片级图片 MJV6文本生成测试 Style raw用法解析

氪學家
1 Jan 202409:16

TLDR本期视频介绍了Midjourney(MJ)新推出的V6 alpha版本,这是自V5.2以来时隔半年的重大更新。视频中对比了MJ V6与DALL-E 3在文本生成图像方面的表现。MJ V6在写实风格图像生成上进行了优化,特别是通过使用Style raw参数,能够生成细节丰富、接近实拍照片的图像。同时,V6在文本生成能力上也有所提升,尽管在拼写准确性上略逊于DALL-E 3,但在图像元素的丰富性上更胜一筹。视频还展示了如何切换到V6模型以及一些新功能的测试。最后,鼓励观众亲自体验V6版本,感受其带来的震撼。

Takeaways

  • 🎉 MJ推出了V6版本,这是一个跨版本的更新,距离上一个版本V5.2已经过去了半年。
  • 🚀 AI绘画领域在这半年内有了显著的技术突破和新产品上市,如DALL-E3和SD XL Turbo。
  • 📈 MJ V6相较于V5.2,在图像质量上有了显著提升,特别是在写实风格图片的生成上。
  • 🔍 V6版本改善了提示词的精确度和一致性,对长文本提示词的理解能力更强。
  • 📌 V6版本提升了图像的提示和混合能力,小幅增强了文本的绘制能力。
  • 🚫 V6测试版不支持某些功能,如方向性拓图、缩放拓图、局部重绘、样式协调器和提示词反求功能。
  • 🔗 官方建议在V6中避免使用photorealistic, 4K8K等提示词,推荐使用style raw参数或较低的stylize值以获得更写实的风格。
  • 🖼️ 通过添加style raw参数,V6版本在生成写实风格图片方面表现出色,细节处理接近实拍照片。
  • ✍️ 在MJ中生成文本时,需要将文本内容置于英文双引号中,V6版本在文本绘制上取得了进步。
  • 🤖 DALL-E3在文本生成的准确性上略胜一筹,但MJ V6在元素丰富性和写实感上更胜。
  • 📚 官方文档提供了关于V6版本的详细更新说明,建议用户亲自体验V6带来的新功能和改进。

Q & A

  • Midjourney V6 alpha版本相较于之前的V5.2版本有哪些显著的改进?

    -Midjourney V6 alpha版本相较于V5.2版本,显著的改进包括更精确的提示词跟随、更长的提示词支持、改善了一致性和模型的知识、提升了图像的提示和混合、小幅文本的绘制能力增强,以及改进了放大器,包括subtle和creative模式。

  • 为什么Midjourney V6 alpha版本不建议使用'photorealistic'这样的提示词?

    -Midjourney V6 alpha版本对提示词更加敏感,官方认为'photorealistic'等提示词为垃圾提示词,因为它们在V6版本中不再需要,V6已经对写实风格图片做了优化,添加这些提示词反而可能导致不理想的效果。

  • 在Midjourney V6 alpha版本中,如何生成更加写实的图片?

    -在Midjourney V6 alpha版本中,可以通过使用'style raw'参数或较低的stylize值来生成更加写实的图片。

  • Midjourney V6 alpha版本在文本生成方面有哪些特点?

    -Midjourney V6 alpha版本在文本生成方面,需要将文本内容置于双引号内,并且对英文文本的绘制能力有所提升,能够生成具有手写感觉的文本图片。

  • DALL-E 3在文本生成方面与Midjourney V6 alpha版本相比如何?

    -DALL-E 3在文本生成方面对提示词的理解能力更强,但在拼写准确度上略胜一筹,然而在出图元素的丰富性上,Midjourney V6 alpha版本更为丰富。

  • Midjourney V6 alpha版本在图片分辨率上有什么变化?

    -Midjourney V6 alpha版本默认生成的图片分辨率是1024x1024,经过放大后,图片的分辨率可以达到2048x2048。

  • 为什么Midjourney V6 alpha版本不支持某些功能,如pan方向性拓图?

    -Midjourney V6 alpha版本是测试版,因此一些功能如pan方向性拓图、zoom缩放拓图、局部重绘、样式协调器和提示词反求功能暂时不支持。

  • Midjourney V6 alpha版本在AI绘画圈中的地位如何?

    -Midjourney V6 alpha版本在AI绘画圈中的地位有所挑战,因为在这半年内,AI绘画领域出现了许多技术突破和新产品,如DALL-E 3和Adobe公司的firefly第二代等,但V6版本的推出显示了Midjourney在提升产品质量和竞争力方面的努力。

  • Midjourney V6 alpha版本在写实风格图片的生成上有哪些优化?

    -Midjourney V6 alpha版本在写实风格图片的生成上,对服装、面部皮肤纹理、毛发细节以及特写镜头的景深关系等都做了优化,使得生成的图片在真实感上有了显著提升。

  • Midjourney V6 alpha版本和DALL-E 3在写实风格图片生成上有哪些差异?

    -Midjourney V6 alpha版本在写实风格图片生成上,特别是在服装、皮肤纹理和景深关系上的细节处理更胜一筹,而DALL-E 3虽然对提示词的理解能力更强,但在真实感的表现上不如V6版本。

  • Midjourney V6 alpha版本在文本生成时,提示词应如何使用?

    -在Midjourney V6 alpha版本中,文本生成时需要将文本内容放在双引号内,以便模型能够识别并生成包含所需文本的图片。

  • Midjourney V6 alpha版本对于AI绘画领域的贡献是什么?

    -Midjourney V6 alpha版本通过其改进的写实风格图片生成能力和增强的文本生成功能,为AI绘画领域提供了新的工具和可能性,推动了AI艺术创作的边界。

Outlines

00:00

📈 Introduction to MJ V6 and AI Art Development

The video begins with a casual introduction to the MJ tutorial series, acknowledging the two-month gap since the last update. The host highlights the release of MJ's new V6 version, an alpha release that marks a significant update from the previous V5.2 version. The host also discusses the advancements in AI art during the past six months, including the introduction of DALL-E 3 by OpenAI, the release of SD XL and SD XL Turbo models by Stability AI, and the emergence of various AI real-time drawing and rendering projects. The video then focuses on the new features of MJ V6, such as improved prompt following, enhanced consistency and knowledge of the model, better image prompts and blending, and the ability to handle small text. The host also mentions the change in prompt sensitivity and the deprecation of certain prompt words like 'photorealistic' and '4K8K', which were previously effective but are now considered ineffective. The video proceeds to demonstrate how to switch to the V6 model within the MJ platform and suggests visiting the official MJ community for more information on the updates.

05:02

🎨 Testing MJ V6's Realism and Text Generation Capabilities

The host conducts a test to compare the realism of images generated by MJ V6 against the previous V5.2 version. Despite the official recommendation against using 'photorealistic' in prompts for V6, the host tests this and finds the results to be less desirable. However, when using the 'style raw' parameter, the V6 version produces highly realistic images, surpassing the V5.2 version. The host explains the 'style raw' parameter, which allows for more control over the output and reduces the AI's tendency to add its own stylistic flourishes. The video also includes a comparison between MJ V6 and DALL-E 3 in terms of their ability to generate realistic images, with V6 showing a significant improvement over DALL-E 3 in terms of detail and realism. Additionally, the host tests the text generation capabilities of MJ V6, demonstrating how to include text within images using quotation marks in the prompt. The video concludes with a recommendation for viewers to experience the new features of MJ V6 firsthand and to look forward to future tutorials on AI art.

Mindmap

Keywords

💡Midjourney V6

Midjourney V6 refers to the latest version of the AI image generation software by Midjourney, which is in the alpha testing phase. It represents a significant update from the previous V5.2 version and is the main focus of the video, showcasing its new features and improvements in image quality and realism.

💡DALL-E 3

DALL-E 3 is an AI model developed by OpenAI, known for its powerful semantic understanding and ability to generate images from textual descriptions. It is compared with Midjourney V6 in the video to evaluate which AI performs better in terms of generating realistic images.

💡Style raw

The term 'Style raw' is a parameter used within Midjourney V6 to enhance the photorealism of the generated images. It is mentioned that using this parameter can lead to more realistic and detailed outputs, as demonstrated in the video through various examples.

💡Text generation

Text generation is the ability of an AI to create textual content within an image, such as 'hello world' written on a note. The video discusses how Midjourney V6 has improved in this aspect, allowing for more accurate and varied text rendering in the generated images.

💡Photorealistic

Photorealistic refers to the quality of images that are rendered with a high degree of realism, resembling actual photographs. The video explains that the term 'photorealistic' is discouraged in Midjourney V6's prompt guidelines, and instead, the 'Style raw' parameter is recommended for achieving a realistic style.

💡Semantic recognition

Semantic recognition is the ability of an AI to understand the meaning of words and phrases in context. It is highlighted as a key feature of DALL-E 3, which allows it to interpret complex prompts and generate images that closely match the textual descriptions.

💡Model iteration

Model iteration refers to the process of updating and improving AI models over time. The video discusses the pace of development in AI image generation, noting that Midjourney's V6 came after a longer gap than previous iterations, indicating a period of significant development.

💡WebUI ComfyUI and FOOOCUS

WebUI ComfyUI and FOOOCUS are user interface programs that are used in conjunction with AI image generation models like Midjourney to enhance the user experience and control over the image generation process. They are mentioned in the context of improving image quality and efficiency.

💡Adobe Firefly

Adobe Firefly is Adobe's AI technology that has been updated to the second generation. It is part of the discussion on the advancements in AI image generation and how different companies, including Adobe, are contributing to the field with their own technologies.

💡AI painting circle

The AI painting circle refers to the community or industry focused on the use of AI for creating artwork. The video script mentions the various technological breakthroughs and new product launches within this circle over the past half year.

💡Discord

Discord is a communication platform where the Midjourney community discusses updates and shares experiences. In the video, the host uses Discord to demonstrate how to switch to the latest V6 model of Midjourney and to access the community's discussions.

Highlights

Midjourney V6版本经过半年沉淀,推出全新功能,旨在提升AI绘画技术。

V6版本相较于V5.2,进行了跨版本的更新,包括更精确的提示词跟随和更长的提示词支持。

V6改善了模型的一致性和知识,提升了图像的提示和混合能力。

V6版本增强了小幅文本的绘制能力,允许在图片中加入清晰的文字。

V6默认生成的图片分辨率为1024x1024,放大后可达到2048x2048。

V6测试版不支持pan方向性拓图、zoom缩放拓图、局部重绘等功能。

V6对提示词更敏感,官方建议避免使用photorealistic, 4K8K等提示词。

使用style raw参数或较低的stylize值可实现更写实的风格。

V6在相同提示词下,使用style raw参数后,出图的真实感远超V5.2版本。

V6版本在写实风格图片上做了显著优化,细节处理接近实拍照片。

V6版本在文本生成能力上有所进步,尤其是英文文本的准确度和多样性。

V6版本在AI绘画圈中与其他大厂产品如DALL-E3、SD XL Turbo等形成竞争。

V6版本的MJ在写实风格和文本生成方面与DALL-E3进行了对比测试。

V6版本的MJ在服装、皮肤纹理、毛发细节和景深关系的处理上优于DALL-E3。

V6版本的MJ在文本绘制上表现出更丰富的元素和更强的写实感。

V6版本的MJ在AI绘画技术方面的进步为用户带来了新的震撼体验。

视频建议观众亲自体验V6版本,以感受其带来的变化和提升。

视频提供了MJ V6版本的测试结果,展示了其在AI绘画领域的创新和实用性。