{"id":3645,"date":"2024-02-04T10:04:06","date_gmt":"2024-02-04T02:04:06","guid":{"rendered":"https:\/\/www.1ai.net\/?p=3645"},"modified":"2024-02-04T10:04:06","modified_gmt":"2024-02-04T02:04:06","slug":"ai%e7%bb%98%e7%94%bbmidjourney-%e5%8f%af%e4%bb%a5%e7%94%9f%e6%88%90%e9%a3%8e%e6%a0%bc%e4%b8%80%e8%87%b4%e7%9a%84%e5%9b%be%e5%83%8f%ef%bc%8c%e5%b9%b6%e5%b0%86%e5%9c%a8%e4%b8%8d%e4%b9%85%e5%90%8e","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/3645.html","title":{"rendered":"AI painting Midjourney can generate images with consistent style, and will soon launch a feature to keep characters consistent"},"content":{"rendered":"
todayMidjourney<\/a>Updated with a new feature called style references that generates images with a consistent style.<\/section>\n

It's also very easy to use, just add the cue word after the\u2013sref<\/strong>Keywords<\/a>, and then just separate links to one or more reference images by spaces, e.g:<\/p>\n

\u2013sref url1 url2 url3<\/strong><\/p>\n

And the weight of each reference image can be adjusted to be enhanced by :: and numbers such as:<\/p>\n

\u2013sref url1: 2:2:3 url3:5<\/strong><\/p>\n

you can also adjust the intensity of stylishing by \u2013sw, default 100 0 to shut down and maximum 1000, e.g.:<\/p>\n

\u2013sref url1 \u2013 sw 200<\/strong><\/p>\n

the \u2013sref feature only affects the image style, so that the previous general image hint (reference to image content) needs to be placed before \u2013sref, and it is still suggested that it be placed at the beginning of the hint (see figure below)\u3002<\/section>\n
\"640-63\"<\/p>\n<\/section>\n
<\/section>\n
Precautions<\/strong><\/section>\n
    \n
  • \n
    The style reference applies to V6 models and Niji V6 models, older models are not supported;<\/section>\n<\/li>\n
  • \n
    Because it is still being tested, the feature may change significantly in the near future, so you need to be careful when using it;<\/section>\n<\/li>\n
  • \n
    If the tendency of the prompt is towards a realistic style, such as modern architecture, and you want to change it to another style referenced, such as illustration, this needs to be added in the prompt as well;<\/section>\n<\/li>\n
  • \n
    Style references have no effect on image cues and must contain at least one text cue word.<\/section>\n<\/li>\n<\/ul>\n
    in the near future, it will be possible to introduce a role-consistency feature, using a similar approach and calling through \u2013cref parameters\u3002<\/section>\n
    Demonstration<\/strong><\/section>\n
    Made several examples using the same cue words, comparing image cues and new style references.<\/section>\n
    Example 1<\/strong><\/section>\n
    Original image:<\/section>\n
    \n
    \n
    cover illustration, Penguin with watercolor, isolated white background, cartoon style, thick line,low detail,no shading<\/section>\n<\/section>\n<\/blockquote>\n
    \"640-64\"<\/section>\n
    Image Tip:<\/section>\n
    \n
    \n
    https:\/\/s.mj.run\/21BbQ-LiGbb4 cover<\/section>\n<\/section>\n<\/blockquote>\n
    \"640-65\"<\/section>\n

    Style reference:<\/p>\n

    \n
    \n
    now, over time, cat with watercolor, collected white ground, cartoon style, pick line, low detail, no shadowing - https:\/\/s.mj.run\/21BQ-LiGbb4-ar-9:11\u2013v 6.0<\/section>\n<\/section>\n<\/blockquote>\n
    \"640-66\"<\/section>\n
    You can see that the difference between the two ways of referencing the images is still relatively large, the cat in the image cue is morphologically close to the original penguin, wearing a hat and scarf, round and chubby;<\/section>\n
    While the stylistic references only use similar styles, watercolors, and lines, the two are very different in terms of image content.<\/section>\n
    That is, the image cue will incorporate screen elements into the newly generated image, while the style reference is more about brushstrokes and texture.<\/section>\n
    To use a possibly inappropriate analogy.<\/section>\n
    Using image hints generates results that will make people say:<\/section>\n
    \"Wow, it's a Monet painting.\"<\/section>\n
    And using a style reference generates results that will make people say:<\/section>\n
    \"Wow, it's an impressionist painting.\"<\/section>\n
    Let's look at another example, this one where the image is more different from the cue word.<\/section>\n
    Example 2<\/strong><\/section>\n
    Original image:<\/section>\n
    \"640-67\"<\/section>\n
    Image Tip:<\/section>\n
    \n
    \n
    https:\/\/s.mj.run\/IGEPdvEpo2M a girl \u2013ar 2:3 \u2013 v 6.0<\/section>\n<\/section>\n<\/blockquote>\n
    \"640-70\"<\/section>\n
    Style reference:<\/section>\n
    \n
    \n
    a girl \u2013sref https:\/\/s.mj.run\/IGEPdvEpo2M \u2013ar 2:3 \u2013 v 6.0<\/section>\n<\/section>\n<\/blockquote>\n

    \"640-68\"<\/p>\n

    Because the image cues and text cues are relatively different, the result is that the former screen logic is not very good, while the latter style reference captures the essence.<\/section>\n
    Example 3<\/strong><\/section>\n
    Example three is awesome, I used the two original images above as parameters, the penguin as an image cue, and the carriage as a style reference, and the effect is as follows:<\/section>\n
    \n
    https:\/\/s.mj.run\/21BQ-LiGbb4 acat \u2013sref https:\/\/s.mj.run\/IGEPdvEpo2M \u2013ar 9:11 \u2013 v 6.0<\/p>\n<\/section>\n<\/blockquote>\n
    \"640-69\"<\/section>\n

    It references both the graphic elements of the penguin figure and the style of the carriage, blending to create kinda interesting graphics.<\/p>\n

    Stylistic fusion<\/strong><\/section>\n
    Because the style reference can support multiple drawings and can also realize the characteristics of style fusion, an example from the Internet:<\/section>\n
    \n
    \n
    photo of a mountain landscape \u2013ar 16:9 \u2013 style raw \u2013 sref https:\/\/s.mj.run\/EDG2Sh3Tktw https:\/\/s.mj.run\/sLXYn-hJJJOW \u2013v 6<\/section>\n<\/section>\n<\/blockquote>\n

    \"640-3\"<\/p>\n

    This is all today, interested partners can actually go to experience.<\/section>","protected":false},"excerpt":{"rendered":"

    Today Midjourney updates a new feature called style references, which can generate consistent images. It's very simple to use it. Sref keywords, then separated by space from one or more links to reference images, e.g. \u2013sref url1 url2 url3 and could be adjusted to the weight of each reference image, enhanced by: : and numbers, e.g. \u2013sref url1::2:2:2:3 url3:5 and can also be adjusted for style strength, adjusted by \u2013sw, default 100, 0 to close, 1,000 to maximum<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[149,144],"tags":[186,1070],"collection":[260],"class_list":["post-3645","post","type-post","status-publish","format-standard","hentry","category-jiaocheng","category-baike","tag-midjourney","tag-1070","collection-midjourney"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/3645","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/comments?post=3645"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/3645\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=3645"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=3645"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=3645"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=3645"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}