Recently usedBean curdIt's really a bit of an uptake, after using so many AI software, beanbags are really riding high on their ability to understand and process images/videos, and the key is that they're free.
Today, we have compiled the most complete and practical P charts on the Internet.directivesJust copy your homework when you use it, and you'll never be able to quit again.
1.change the background
The key to getting rid of the sense of pseudo-humanity is to emphasize that only the background is changed, not the characters
Only change the background to the living room of the house and keep the characters the same
If you have a very clear preset for the background to be replaced, you can also describe it in more detail, the clearer the description, the closer the final picture is to what you expect in your mind!
left instruction
Only change the background to the living room at home, behind the living room is a large floor-to-ceiling windows, outside the window is the Shanghai Oriental Pearl Tower, to keep the characters unchanged!
right-hand command
Only change the background to the living room at home, behind the living room is a large floor-to-ceiling windows (23-story building looking out of the field of view), the window is the Shanghai Oriental Pearl Tower, to retain the same characters
The image on the left is the first time I generated it, I found that the bottom of the Oriental Pearl appeared on the image, I wanted the window view to be a relative high-rise view, so I optimized another version of the cue word to add the height of the building, and the effect was more in line with my expectations
You want to add any details to describe the content is possible, and there is no so-called formula, as long as the description is clear on the line, just like people chat, the clearer the expression, the more the AI can get.
2.Optimization of light and shadow
Light and shadow is a very important influence in the photo, but most people are not very good at adjusting light and shadow, the parameters are too complicated. At most, you can use the one-click filter that comes with the Ptu software.
AI is very good at doing this
Adjust the light and shadow of the photo to emphasize the people more. Keep the characters the same, only change the light and shadow
It should be emphasized that the characters can't change, and when they do, the sense of pseudo-humanity comes again
Of course, if you're not a pure geek, you can also describe the way the light and shadow are adjusted in a bit more detail
Adjust the light and shadow, strengthen the contrast between dark and light to highlight the characters, the first morning sunshine on the face of the feeling. Characters remain unchanged, only change the light and shadow
Once the light and shadow are adjusted, the sense of seniority immediately comes out isn't it!
3.Creative Synthesis
Using beanbags is still proper emoji magic, generating 10 of them for you in a minute doesn't matter. And the picture with the text is very grounded. Text and figure echo is also very good!
A different style of drawing would be fine, the cartoon style is literally AI ruled territory!
Crossdressing doesn't matter either!
But remember all the manipulation of the person, now you need to talk to him and keep your face the same!
Please generate photos of him in Zhuang, Miao, and Uyghur costumes. The face remains unchanged
Local fine-tuning is also a more common need in our daily P charts, and the beanbag can be covered off as well! For example, localized switching action, not too silky smooth ow!
A breeze blows through, and the hair whips and
Give the character a different action, coordinated movement, face unchanged
4.Motion Graphics
Just give Beanbag the original image and tell him how to generate a video for you, and what you want the motion effects to look like
Of course it's okay for the original image to be AI generated!
Help me generate a video: a breeze blows through the hair and flicks it
5.Don't have a Pic idea? Let Beanbag give you advice
Let him plan your program first:
I want to make a cyberpunk style poster with the characters in this photo, what do you think I can do to help me plan it out
Determine the program and then have him generate the images directly for you:
6.shortcoming
1. Portrait retouching is not natural enough, and you can tell at a glance that it's AI-processed
Maybe it's because the human body is so delicate, and Nuwa didn't just make people up 😄 AI does have a lot of room for improvement in this area, especially when comparing it to a real person, the gap is obvious.
I tested a couple scenarios to show you the counter case:
Only acne and dark circles are removed from the face, and lines are removed from the face. Skin should keep the natural texture without stiffness, retaining the details of skin pores and presenting a real sense of texture. All other things remain the same, do not have an unrealistic feeling
It's useless to keep it real.
2. AI only recognizes "normal" and has limited understanding of special designs
The AI automatically "fixes" what it thinks is abnormal.
For example, when he was asked to dress Labubu in a Hmong costume, he ended up "fixing" Labubu's classic big mouth...
Labubu without a mouth is really weird 😅😅😅😅😅😅😅
Dress labubu in a Hmong costume and headdress, everything else stays the same.
But then again., despite all these glitches, it's totally flawed! It's so powerful and free, what do you need a bike for 🚲
And I would assert thatHumans have underdeveloped the beanbag 1%The first thing you need to do is to get your hands on some of the most useful tips you can find! In addition to Pics, there are so many useful tips waiting to be discovered~