ComfyUI Relighting ic-light workflow #comfyui #iclight #workflow - 雙語字幕
Hi.
Today will get to know a new and very interesting tool in my opinion,
which allows us to change the lighting on a certain object, while adjusting colors and shadows, and also referring to the background.
The project and all the credit goes to Lyisville,
and I will of course leave a link to the repo in the description of the video.
Here is a demo where you can experiment, and understand a little how this tool works.
And also a big thanks goes to Kijai who made this project available.
I will also leave a link to that in the description.
You can download the necessary nodes from the comfy manager,
and perhaps the simplest way is to load the workflows attached to the examples and simply install the missing nodes.
The models can also be found in the manager,
and you can download them manually from this page, and simply put them under the unet folder under the models folder.
So let's take this workflow from the Kijai example and see what we have to play.
The red node is probably a node from another package, and according to the title it is responsible for disconnecting an object from the background.
So instead of installing it, I will simply replace it with remove background which I already have on my computer.
Let's start by loading an image, and here we need another image, probably for the background.
I want to put the model underwater so we'll load this image for the background.
I change the model to the model that exists on the computer, and in the prompt we will write a picture of a woman underwater.
Q is pressed,
and as you can see,
a mask was created for us here,
which probably has an important role in the process,
and we got a not-so-good picture, and especially distorted in terms of proportions, so first we'll fix that.
I change the key proportion to true, and the amount of pixels to 768 lengthwise and widthwise, and press Q again.
You can see that the proportions are now correct, but the quality is not so good.
And from the attempts I made, it seems to me that it works better, without adding a background image.
So let's check it out.
Notice that there is a question mark here.
If we click on it, you can see that in order to add a we need to work with a certain model.
Right now we only want to work on the character,
so I disconnect the latent of the background, and choose the appropriate model and press Q.
And now you can definitely see the effect the mask has on the process.
And the image feels more harmonious compared to the previous version.
Let's try to move the mask a little and see how it affects the result, so that we don't have to wait long.
In order to see the change in the mask, I'll put the sampler on mute for a moment.
This process is a bit annoying,
so we'll soon arrange that it will be more convenient for us to work,
but let's first see how the change in the mask affects, and you can definitely see that the lighting now comes from the right side.
To optimize the process of creating the mask, I want to be able to draw on the image.
which will allow me a little more freedom and comfort.
Let's change the workflow a bit.
Instead the create mask node, we will use load image to draw the mask where we want on the image.
What we now need is just to adjust the mask to the size of the image after the resize.
Then we will look for the resize mask node,
and the length and width will be changed to input, and we will connect them to the resize image.
node.
And we will connect this mask to the grow mask with blur.
And you can see how the lighting changes depending on the position of the mask.
Another thing worth playing with that affects the result is the blur radius.
And as you can see in Lyisville's demo,
the prompt and the description of the lighting also The problem is that the picture is a bit dark for me,
and I want to get some of the atmosphere in this picture,
which we tried to connect to the background before, for that I will also connect an IP adapter, and we'll see what we get.
Notice that the weights of the IP adapter are not too strong, compared to the change in lighting you want.
You can also play a little with the start at and end that to moderate the effect of the IP adapter.
Let's try with another picture.
Let's say we want to put this robot on the street at night,
and that it will be lit a little on the right side,
so I drew the mask where I want the light, and I'm uploading a reference image of the IP adapter for the background.
Don't forget to change the prompt to that.
And this is the result we got.
If you notice the transition between the robot and the background is not smooth,
and it is probably related to remove background, let's try to change the transparency to false.
And now the transition is much smoother.
I lower the blur radius a bit, and raise the weights of the IP adapter a bit.
And, perhaps one last thing, let's see how the robot is received without the IP adapter.
Here we have the IC light conditioning,
which allows us to play a bit with the multiplier,
which in terms of the result is more similar to raising the contrast, and now we will just add the IP adapter.
And this is the final result.
In short, a very interesting tool, which definitely adds control to the creative process.
So I hope you learned,
and that we will meet in the next lessons,
of course you are welcome to subscribe to the channel, ask questions and like if you liked, and most importantly, have fun!
解鎖更多功能
安裝 Trancy 擴展，可以解鎖更多功能，包括AI字幕、AI單詞釋義、AI語法分析、AI口語等
兼容主流視頻平台
Trancy 不僅提供對 YouTube、Netflix、Udemy、Disney+、TED、edX、Kehan、Coursera 等平台的雙語字幕支持，還能實現對普通網頁的 AI 劃詞/劃句翻譯、全文沉浸翻譯等功能，真正的語言學習全能助手。
支持全平臺瀏覽器
Trancy 支持全平臺使用，包括iOS Safari瀏覽器擴展
多種觀影模式
支持劇場、閱讀、混合等多種觀影模式，全方位雙語體驗
多種練習模式
支持句子精聽、口語測評、選擇填空、默寫等多種練習方式
AI 視頻總結
使用 OpenAI 對視頻總結，快速視頻概要，掌握關鍵內容
AI 字幕
只需3-5分鐘，即可生成 YouTube AI 字幕，精準且快速
AI 單詞釋義
輕點字幕中的單詞，即可查詢釋義，並有AI釋義賦能
AI 語法分析
對句子進行語法分析，快速理解句子含義，掌握難點語法
更多網頁功能
Trancy 支持視頻雙語字幕同時，還可提供網頁的單詞翻譯和全文翻譯功能