I've noticed a lot of peeps asking the same question lately?Prompt wordDo not know how to write? And want to make the same figure as others, I believe this is also a lot of newbies common problems~
Today, I'm going to teach you an easy way to do this:From the pictureextrapolate backwardsCue word, when you see those explosive AI works or beautiful pictures on the Internet, you can directly screenshot down the picture to do the backpropagation.
So what is backpropagation: backpropagation prompt word meaning let AI language big model, after recognizing a picture, extract the descriptive information in its picture, such as style, subject, action, scene and details. With these words we can in using AI to generate the same similar original pictures.
Let's see the demo first (the first original image, the last four with different backpropagation effects)
step by step
The steps are actually broken down into three:
Step 1: Once you get the reference map, give the AI Big Model a request command to get the cue word for the inverse.
Step 2: Modify the cue words to suit your needs.
Step 3: Open Imagine Dreams and type in the prompt word to come up with a picture.
I won't focus on how to get the reference image, when we see other people's videos or have a reference image we can just take a direct screenshot.
Requirement commands, a few of which are summarized here for you:
The easiest: just 2 words, replica
at presentBean curd[Image Generation] [Super Creative 2.0] In the gray-scale test, the main one to use the comfort of free.
Upload a reference image, enter the model and enter the replica, there is no need for most of the language or to ask for details such as style + subject + visual elements (composition, colors and tones, light and shadow, texture and texture)
I'm sure this is a step that anyone with a hand would know, so here's a compilation of some commands for you to use in reverse.
The following demo has found a former model.
Simple Version:
Group 1: Help me analyze this picture and write the prompt words for the instant dream AI text born picture to facilitate the generation of the same picture
rendering (computing)
Group 2: Help me describe this picture in detail, organizing it into a paragraph for the prompt words for the Vincentian picture
rendering (computing)
Advanced Edition:
Group 3: Please parse this image to generate a textual cue that will guide the AI tools in recreating a similar work. The cue must cover the following key elements: the core subject of the image, the temporal and spatial setting of the scene, examples of artistic style references, color palette, compositional perspective, and any other additional details that need to be emphasized.
Group 4: Detailed disassembly of this picture, you need to describe the picture in words from the main content, scene setting style reference, tone and color, composition perspective, details to supplement these perspectives, and summarize to generate a text that can be used in AI mapping tools to generate a raw picture prompt words.
rendering (computing)
Here's a raw graphic cue word to share with you below
Outdoor badminton court scene, youthful girl subject, black long hair, wearing white short-sleeved sports top (with light blue splicing, subtle texture), white sports skirt (black brand logo on the side of the skirt), wearing pink and white gradient sports shoes, white calf socks (socks with small pink letters), holding light blue badminton racket with both hands, one leg elegantly lifted up to show a playful gesture. Scene for the green fence, red and green field of the real badminton court, the background moderately blurred to create shallow depth of field. Style reference fresh Japanese youth photo, color tone with white, light blue, pink and white as the main, with green environment color and red field color, composition for the side of the middle view, highlighting the youthful vitality, sweet and healthy atmosphere, emphasize the texture of the dress, brand logo, shoes and socks details and natural posture, eyes and camera interaction.
Group 5: Please analyze this image and generate a detailed text-to-graphic prompt describing the style/art form + describing the subject + layering visual elements (composition, color and tone, light and shadow, texture and grain) + setting + atmosphere + detailing + quality requirements. The prompt you provide must be a full paragraph in Chinese, and should be able to guide the AI mapping tool to recreate a similar piece.
rendering (computing)
Of course we don't have to be in a beanbag, e.g., Tencent Yuanbao, basically all the big AI models are available for it to backpropagate the cue word. The same upload reference chart, just enter the appeal instruction, or directly enter the backpropagation prompt word.
Finally, I put together a comparison between the original image and the image generated a few times after the cue word was backed out, so you can see the effect.
Learn to use these backtracking applications and you too can write accurate prompt words, so go ahead and give it a try.
Test points: the cue word after the back-propagation does not say 100% consistent, sometimes individual style or action angle is out of the way, need to optimize their own in the manual continuous tuning.
Practical difficulty: ★✰✰✰✰
Tools and methods have the last is also a test of perseverance and creativity, the lack of one can not~
That's what I'm sharing today.Everyone can get hands-on, easy to learn, hurry up and try it~!