플럭스 1: 혁신적인 AI 이미지 생성 모델 공개

### AI 이미지 생성의 혁신: 플럭스 1을 만나보세요
* 🌟 현실과 구별할 수 없는 이미지를 생성할 수 있는 **Flux 1이라는 새로운 인공 지능이 3천만 달러 이상의 자금을 지원**받아 등장했습니다.
* 🖼 이 모델은 이전에는 AI가 처리하기 어려웠던 **손과 같은 복잡한 이미지의 생성을 크게 개선**합니다.
* 💻 Flux 1은 속도가 빠르지만 품질이 낮은 버전, **고품질의 Flux Pro, 중간급 모델 등 세 가지 모델**을 제공하며, 모두 무료로 사용할 수 있습니다(Midjourney와 달리).
* 🚀 플랫폼인 블랙 포레스트 랩은 팀과 모델에 대한 자세한 정보를 제공하며, **미드저니 버전 6보다 우수한 점**을 강조합니다.
* 🎨 사용자는 **특정 텍스트와 사실적인 디테일이 있는 이미지**를 생성하여 손 변형과 같은 일반적인 AI의 함정을 피하고 사진처럼 보이는 결과를 얻을 수 있습니다.
* 🔍 Flux의 오픈 소스 특성 덕분에 **기존 AI 이미지 생성 플랫폼에 통합**하여 접근성과 기능을 확장할 수 있습니다.
* 💡 Glif와 같은 플랫폼을 통해 Flux Pro를 활용하면 사용자는 **고급 기능을 잠금 해제**하여 자세한 안내와 함께 **고품질 이미지**를 제작할 수 있습니다.
* 🎉 전반적으로 **Flux 1은 창의성을 민주화**하여 누구나 자신의 아이디어를 표현하고 전문적인 결과물을 통해 비전을 실현할 수 있도록 지원합니다.

Flux 1은 거의 구별할 수 없는 AI 생성 이미지를 무료로 생성하여 이미지 생성에 혁신을 일으켰습니다.

It seemed that we had seen everything in image generation by artificial intelligence, especially with the latest updates from platforms like Midjourney. However, that's not the case. With a funding round of over 30 million dollars just a few days ago, a new artificial intelligence has emerged that is capable of generating images that are practically impossible to differentiate from reality versus those created by artificial intelligence. Until now, creating images such as hands was difficult when we wanted to achieve good results, but now we will be able to obtain significantly better results. In fact, these three models are established as the best currently available. But the best part is that, in addition to having ownership and being able to create uncensored images, it is also an open model that, unlike Midjourney, we can use completely free of charge. Sounds good? Well, let's get to know it. The new artificial intelligence that we will be looking at is touted as the new era of creation, and it does so under the name of Flux 1, which is this model here. Before continuing, I remind you that all the links and information will be available below in the description. Once you click on the first link, you will land on this page here, where we can find this new model from the company Black Forest Labs. Before anything else, if we click on this section, we will find an article where we can gather more information about the team behind it, data such as the funding round I mentioned earlier, and above

models are the best, having two in the first position and their smaller model in the fifth, but all of them ahead of my Journey version 6. They have indeed launched three models at once. If we scroll down a bit, we will find these three models here. Each of them has its peculiarities, as the first one will allow us to create images very quickly, although the quality will be slightly lower compared to the other two larger models. In the other one, we have the Flux Pro, which offers the best possible quality. The one in the middle is precisely the midpoint in terms of speed and quality. In this video, we will be testing all three models. Although theoretically, the Pro model is paid, I have found a trick that will allow us to use it for free to see a bit of its results. I will go back to the homepage, and if we scroll down a bit, we will find some results that are very good, like this one here, where we can see all the textures. Although the image is quite creative, both the play of light and shadows and all the textures and colors are perfectly created. All of this has been done with a very simple prompt, using just this phrase here. We will also be able to see its capabilities to create images with specific texts, as specified here for the text to appear on the screen that says Flux. However, beyond a single word, we can find other more complex integrations, like this one here

también frases más largas que están perfectamente integradas con todo el contexto en el que aparecen. Aunque hay algo por lo que este modelo destaca, es en todas las imágenes realistas que podremos crear, encontrando algunos ejemplos que serían difíciles de diferenciar si están creados por inteligencia artificial o si realmente son fotografías. Pero, sobre todo, tiene en cuenta los detalles, como evitar la deformación de las manos u otras características específicas del cuerpo humano, como sería el ejemplo de este ojo. Todos estos ejemplos suenan genial, y de hecho, si hacemos scroll hacia abajo, vamos a encontrar diferentes accesos a cada uno de sus modelos. Aunque, como te comentaba al inicio de este vídeo, al ser un código abierto, cualquier plataforma que ya esté consolidada en la generación de imágenes por inteligencia artificial podrá integrarlo en su plataforma. Así que, más allá de esto, vamos a poder utilizarlas en plataformas como la de Cart, una plataforma muy similar a la de Leonardo AI, que normalmente trabajaba con modelos de Stable Diffusion, pero que recientemente ha integrado también este nuevo modelo de Flux para poder utilizarlo. Le daríamos a iniciar sesión y ahora, desde aquí, podríamos registrarnos con cualquiera de estas opciones. Aunque yo, como siempre, lo haré con Google, con la cuenta de todos los tutoriales. Una vez hecho esto, ahora sí, clicamos en este modelo de Flux y desde aquí podremos poner nuestro prompt, como por ejemplo, un hombre surfe

It tells us that we are using its fastest model, and in fact, you can check it here, as I am speaking, it is finishing loading. However, you should keep in mind that this is the one with the lowest quality, but you will see that it is still quite good. Perfect, we already have this first image, which is honestly quite good, although it is true that the Sur table has been placed on its back, making it look like a Ninja Turtle. Beyond this, we can focus on all its details, and the truth is that it is great, having all the lighting effects perfect, making it seem like the sun is in this part and casting all these shadows according to the shape of its face. Even if you look at this shoulder, we have the shadows exactly matching the hairstyle of this person. The truth is that it is a model that takes into account many details. Here I just created another example where I specified a little better that it is waving with a raised hand, and it gives us this result here, where it would indeed look like it is surfing. And again, all these details are present, having both this hand and the one up here, which is practically perfect, but it seems that this area here is a little off. But in general, the result is not bad at all. Beyond this, the images can also contain specific texts, and to test it, I will erase here, and now I am going to write that we will

Imagine that you can achieve a very good result on the first try in all the tests I am showing you, having this sign with the letters perfectly integrated saying Alex Javi and, ultimately, a lion surfing on the beach. Here, I have also created a girl showing her feet so we can see how she does it perfectly, with no deformation at all. We can see the five toes that all people have. Similarly, I also requested another image of a man's feet, which turned out quite well on the first try. Other examples include feet on the beach, where we can find this example. Again, all the details are great, appearing as if it is a photograph. As I mentioned at the beginning of this video, the images we want have no censorship. Here we can find an example that is obviously censored so that my channel doesn't get taken down, but where we have a naked girl. Here we have another example, in this case of a man, where I specified that he should be sitting in the middle of the street on a sofa, and another girl who also came out perfectly on the first attempt, where I used exactly the same prompt as the previous girl I showed you. What I just showed you really cannot be done on any platform since two circumstances must be met: on one hand, the model must allow you to create uncensored images, and on the other, if you do it through the internet, the platform containing that model must also allow that.

Images from DALL-E must also take into account that they have payment plans, although their free version is more than sufficient, as it provides us with 150 points daily with which we can create different images. However, if you wanted these models, you could also install them locally on your computer or open them from servers like Hugging Face. From there, if I input a prompt exactly like we did in the previous platform with the lion holding the sign, if I click Run, you can see that it also processes very quickly, and in about 5 seconds, it has created this image, which is really great. Although it is true that it has rushed a bit and the text isn't quite right, since it works so fast, in another 10 seconds, I have this other image here, which is fantastic. If I like it, I could click the download button and automatically have it on my device. Everything we just tested was with the smallest model, which theoretically has lower quality but yields faster results. Now, we will be testing the intermediate model, which has slightly better quality. For that, you will also find this other Hugging Face space in the description, where if we input the exact prompt we used before and click Generate, it will indicate that the time to generate the image will be a bit longer. However, as you can see, it is still quite fast, and here we have the result from the intermediate model, where the first one has given us

Surfboard with the name Alex Javi, and we can really appreciate even greater quality. All these images are great, but be careful when comparing them with other models. Since this one is not a platform but rather a model that integrates into other servers, it is limited to creating images based solely on the descriptions we provide. However, if we look at other models like DALL-E 3, which, by the way, for those of us with the free version of ChatGPT, we can now create images from here. If I instruct it to create exactly the image we just made in the platform and click send, it would automatically generate this image for me in the free version of ChatGPT that I have. Here we would have another result that is very good, as it is quite a bit more creative. But be cautious because there is a trick; if we click on the information symbol, we can see that it has done a prompt engineering according to what I was asking. But we can also integrate this into Flux. To do this, it would be as simple as clicking here at the top right to customize ChatGPT, and from here below, I will put this prompt that I will leave in the description, where I am basically telling it to apply much more detailed descriptions to create a better image for when we do it in the Flux model. I enable this for new chats, and when I save it, if I now click here to start a new chat and input exactly the

I copy and paste it, for example, here in the fastest model. When I click Run, I would already have this other image with more vivid colors in accordance with the entire description we have provided here. Now, if we want to unleash its full potential, we need to use its more advanced version called Flux Pro. This version is paid, but we can actually use it through a platform called Glif, which is this tool here that truly focuses on allowing us to create any artificial intelligence easily. 

First, we need to register by simply clicking on this button, and here I will do it again with Google. Once we have that, we click on Build, and it would be as easy as clicking the plus sign. From here, we will be able to use different artificial intelligences and triggers. However, what we are interested in for this video is to click on the image generation section, as if we go to this selector, we will find the Flux Pro model. We will click on it, and among other settings, keep in mind that you will also be able to choose the aspect ratio. Although I will leave it again at one by one, and from the advanced settings, you will also be able to modify how many steps you want the creation process to take. 

Keep in mind that the more steps you choose, the better quality you will have, but it will take longer to generate. To create our image, you simply describe here what you want to create,

of a small kitten on a finger. So, I select the prompt, copy it, and paste it onto this model. With this, we simply hit RAM, and here we have this other image where we can see much higher quality compared to the previous models we have tested. The textures on the kitten really show super high quality, but especially if we look at the hands. Besides being quite well done, we have all these marks that appear in a photograph when we take a close-up of hands. To give you a clearer comparison between Flux and MidJourney, I prepared a comparison with four images that were already made in MidJourney. I used both the smallest model and the pro model, and I only took their first result. If we compare it with MidJourney, these would be the results. 

The results are not bad at all, and finally, we can do it with a free model. Now, if we want to get the most out of it, I’m going to share a very good trick. If we return to this platform, beyond just writing the prompt here, we will also be able to simulate what we did earlier with ChatGPT, where it described in much more detail what we wanted to create. Doing this is very simple and automatic from this platform. To do this, it would be as easy as putting a block of text in front; here I will change the variable to text, and I will title it prompt. Now, below, I will add another

Rational conversation. So, we are going to click this button, and from here, notice that we have this button, which is exactly the variable we are taking from here. So, what we are going to do is put exactly the same prompt that we used before, only here we will tell it to do it based on the description, and in quotes, we will put the variable that is what the person, in this case, I, will input for the image we want to create. 

So, the flow would be as follows: Here we will put the text, which will be processed by this artificial intelligence to make it much more detailed. If we click on advanced settings, we will be able to select another model, such as GPT-4 or Mini. Right below, it is very important that we delete this prompt that was written before and select the result from the previous block, which I will actually rename to optimized. So, here I will click on optimized, and now I will test it with something unusual, like a lion giving a TED talk. If I hit run, we will see that automatically the text is converted to optimized text. In fact, if I scroll down a bit, we will see that the prompt it created for me is this one here, and with this, it is automatically generating the perfect image. Here we would have our lion on a stage, and if we go to the detailed description, we would find the characteristics specified in

Now, keep in mind that this model, beyond creating images, is also capable of editing them. For this, we will need to use the Flux model, specifically inpainting, from where we will be able to upload an image in this section here. So, I will drag the first image we created in this tutorial directly here, and now from here we will have this mask symbol. So, I will click on it, and we will be able to select a size; I will leave it around here. Now, from here, we will be able to paint what we want to add. In my case, as you can already imagine, some sunglasses. So now here in the input field, I will type sunglasses, and when I click submit, it will tell me that the generation will take less than a minute. By now, we would have the person we created earlier with the previous artificial intelligence, in this case, with sunglasses that, while not dark, have integrated perfectly, even respecting the light areas we had in the image before. Well, what do you think about being able to do all this for free? Let me know in the comments and share if you liked it. As we can see, the most powerful artificial intelligence models are becoming increasingly accessible, allowing us to achieve professional results easily. These types of innovations truly democratize creativity, making it possible for you to express your ideas and turn everything you can imagine into reality.

Como Midjourney pero… GRATIS, MEJOR y SIN Censura 🔓 Tutorial y trucos con FLUX.1

Alejavi Rivera

플럭스 1의 등장: AI 이미지 생성의 판도를 바꾸다

3천만 달러가 넘는 자금을 지원받아 현실과 거의 흡사한 이미지 생성을 제공하는 새로운 AI인 Flux 1이 등장했습니다.

속도와 품질이 다른 세 가지 모델을 제공하며, 모두 무료로 무수정 이미지 생성이 가능합니다.

플럭스 1은 미드저니의 버전 6을 능가하는 업계 최고로 평가받고 있으며, 프로 모델을 무료로 테스트해 볼 수 있는 특별한 기회를 제공합니다.

플럭스 모델이 이미지 생성에 미치는 영향

Flux 모델은 사진과 구별하기 어려운 사실적인 이미지를 생성하는 데 탁월하며 인체의 특정 디테일에 주의를 기울입니다.

오픈 소스이기 때문에 기존 AI 이미징 플랫폼에서 쉽게 통합할 수 있습니다.

레오나르도 AI와 유사한 플랫폼인 Cart는 Flux 모델을 통합하여 사용자가 이를 등록하고 다양한 프롬프트가 있는 이미지를 생성하는 데 사용할 수 있도록 합니다.

AI로 완벽한 이미지 결과물 얻기

AI 모델은 첫 번째 시도에서 사자가 서핑하는 모습이나 완벽하게 묘사된 발과 같은 독특한 프롬프트를 포함하여 매우 섬세하고 검열되지 않은 이미지를 만들 수 있습니다.

DALL-E는 이미지 생성을 위해 매일 150포인트가 제공되는 무료 버전을 제공하며, 로컬 설치 및 허깅 페이스와 같은 서버 옵션도 사용할 수 있습니다.

모델의 속도는 다양하며, 가장 작은 모델은 빠른 결과를 제공하고 중간 모델은 약간 더 나은 품질을 제공합니다.

AI로 이미지 만들기: 단계별 가이드

알렉스 제이비 서핑보드 모델은 뛰어난 이미지 품질을 제공하지만 기능이 제한되어 있습니다.

무료 버전의 ChatGPT를 통해 제공되는 DALL-E 3는 사용자 설명을 기반으로 이미지를 생성할 수도 있습니다.

이미지 생성의 잠재력을 극대화하기 위해 사용자는 ChatGPT의 프롬프트를 사용자 지정하고 고급 AI 이미지 생성을 위해 Glif 플랫폼을 통해 유료 Flux Pro 버전을 사용할 수 있습니다.

손가락에 새끼 고양이 이미지를 생성하는 데 사용된 모델은 이전 모델에 비해 텍스처와 디테일이 크게 향상되었습니다.

Flux와 MidJourney를 비교한 결과, 무료 모델을 사용하더라도 Flux의 결과물이 상당히 경쟁력이 있음을 알 수 있습니다.

이 플랫폼은 텍스트 블록을 사용하여 자세한 설명을 제공함으로써 이미지 생성을 향상시키는 트릭을 제공하여 최적화된 프롬프트와 더 나은 결과를 얻을 수 있습니다.

플럭스 모델로 이미지 편집하기

Flux 모델은 이미지 편집을 위한 인페인팅 기능을 제공하여 사용자가 이미지를 업로드하고 수정할 수 있습니다.

사용자는 마스크 기호를 사용하여 크기를 선택하고 선글라스와 같이 추가하려는 영역을 칠할 수 있습니다.

이 모델의 강력한 기능에 대한 접근성이 향상되어 창의력을 대중화하고 사용자가 손쉽게 전문적인 결과물을 얻을 수 있게 되었습니다.