Microsoft Bing 이미지 크리에이터: 기능 및 사용자 환경

Microsoft의 Bing 이미지 크리에이터는 텍스트 프롬프트에서 이미지를 생성하여 인상적이지만 때때로 결함이 있는 결과를 보여줍니다.

### Microsoft, 빙 이미지 크리에이터 공개: AI로 생성된 예술의 새로운 시대
* 🖼 Microsoft는 **텍스트 프롬프트를 이미지로 변환하는 생성형 AI인 Bing 이미지 크리에이터**를 출시하여 Bing 채팅 서비스의 기능을 확장한다고 발표했습니다.
* 🔍 이 외에도 구글은 현재 미국과 영국 사용자를 대상으로 실험적으로 출시되었지만 아직 한국에서는 사용할 수 없는 **제너레이티브 AI인 바드(Bard)**를 소개했습니다.
* 🌍 비영어권 국가에서는 언어 정확도 및 잠재적 오류에 대한 우려로 인해 **제한적으로 제공**되므로 보다 발전된 언어 모델부터 시작하는 것이 현명합니다.
* 💻 Bing 이미지 크리에이터의 사용법은 간단합니다. 사용자는 모든 브라우저를 통해 액세스할 수 있으며, 이미지를 생성하려면 **영어로 메시지를 입력**해야 하며, **테스트용 이미지는 하루에 25개로 제한**됩니다.
* 🐶 크리에이터는 **우주복을 입은 시바견의 매력적인 디지털 아트 작품** 등 다양한 사례를 소개하며 AI의 기능과 개발 타임라인을 강조합니다.
* 🎨 사용자들은 **자전거를 타는 악어와 먼 곳을 응시하는 여성 등 인상적인 결과와 실망스러운 결과**를 모두 보고했는데, 디테일과 정확성에서 약간의 결함이 있는 것으로 나타났습니다.
* 🖌 이러한 문제에도 불구하고 AI는 **반 고흐와 같은 유명 아티스트와 일본 만화 및 픽셀 아트**와 같은 스타일에서 영감을 받은 이미지를 생성하여 예술적 스타일을 성공적으로 캡처합니다.
* 🍣 전반적으로 **Bing 이미지 크리에이터는 창의적인 비주얼을 생성하는 데 상당한 잠재력**을 보여주며, 향후 업데이트를 통해 **구글의 바드에 대한 더 많은 탐색**이 이루어질 것으로 예상됩니다.

Hello, this is Kim Ha-gyeong. Last night, Microsoft announced another generative artificial intelligence called Bing Image Creator. Since it deals with images, it will be an AI that generates images. Previously, with Bing Chat, when you input text, you would receive text back. Now, when you input text, you get a picture or image instead. The images that show the background here are the ones generated by Bing Image Creator. 

Additionally, there was another announcement last night. It seems like these companies are working non-stop, perhaps even 69 hours a week. Google announced a generative AI called Bard. They had previously announced it, but this time they made it publicly available to the general public. However, it is currently being released experimentally. So, it is available in the United States and the United Kingdom, but it is not yet supported in South Korea.

I know that this is not possible in most countries, with the exception of the United States and the United Kingdom. There are various reasons for this, but last time, during an event where this was announced, there were some issues when incorrect answers were presented. Because of that, especially in countries where the language, like ours, is less developed, there is a greater possibility of errors and potential problems when launching this feature. So, I think it makes sense to start with English-speaking countries where the language model is more developed. I wanted to apply for this as well, but it’s not available in our country. Therefore, I asked my nephew in the United States to help me out, and I tried a few things, which I will share later. Today, I will only talk about the Bing Image Creator. Using the Bing Image Creator is very easy; you can use any browser.

Now, I'm currently using Chrome, so I've opened Chrome. You can also use Edge; it doesn't matter. You just need to type this into the address bar. After typing it, hit enter, and the screen will change like this. This screen will appear, and you will see the prompt input box. It says, Describe the image you want to create. Here, you can type what you want to draw. For example, if you want to create a pancake, you can type something like, Make a six-layer pancake. However, one important point to note is that you can only use English here; Korean won't work, so keep that in mind. If you want to use this more effectively, you can click on Sign Up to register and create an account.

If you have a Microsoft ID, you can log in and start using it. Once you log in, a window like this will appear. In this window, there is a prompt section where you can type in exactly what I mentioned earlier. Then, images that match the text you entered will be generated. It seems that you can create up to 25 images per day for testing. I have already tested 17 images, so I only have 8 left. Below this, there is a section called Idea Exploration, where Bing seems to have gathered some decent examples it has created before. If you click on the Creations section, you will see a list of what you have created. You can check it later. Among them, there is a cute one with a Shiba Inu, and below that, there is what seems to be a hamster.

I clicked on this, and when I clicked on the hamster, this image popped up. Next to it, it showed Q4G Hamster: Folding of Light Digital Art. It's a very cute and charming hamster holding a glowing orb. It seems that this image was created about 28 days ago. Below it, I noticed a Shiba Inu, which is an astronaut Shiba dog. Anyway, it has a red and white spacesuit, and the background is yellow. I requested it to be drawn in digital art form, and this is what came out. This picture was actually made five months ago. What I'm saying is that just a few days ago, the Air and Extraction PT was completed last August, and after finishing tuning tests and such, it has recently been released.

It was released about six months later, but the fact is that it has already been prepared for almost five to six months prior to that. I can tell this because I checked a few images, so it could have been ready even longer. That's all I have. So, I tried a few different things, and when I asked for an illustration of a crocodile riding a bicycle in the bamboo forest, it was drawn like this. However, there’s something peculiar about it. The crocodile has its legs on the pedals, but its front legs are too short, so it's resting its chin on the handlebars. In another image, the front legs are holding the handlebars while the back legs are on the saddle. In this way, the shapes are drawn quite well, but I noticed some disappointing details here and there. I wasn't late to this, but there was a man smoking a cigarette.

I had to draw a man smoking a cigarette in black clothes, and it turned out well. However, he still has four fingers, and there are some instances where the fingers overlap. I mentioned it was a man smoking a cigarette, but the cigarette appears to be floating, not connected to his hand or mouth. There was also a request to draw a photo of a woman gazing into the distance, but her expression looks strained, and one of her eyes is looking to the right while the other is looking to the left, creating a strange gaze. There are some disappointing aspects. After that, I asked to draw the same image in the style of Picasso, and it was drawn in that style. The next request was to draw it in a Japanese animation style.

When I asked for a drawing, it was created in a Japanese manga style. Then, when I requested a pixel art style, it was drawn like this. It gives off the vibe of the old Hercules game console. In this case, it seems that it has learned from such games, as I could see the text appearing above and below, just like in those games. 

In addition to this image, I tried a few other things. One was to have Seoul drawn in the style of Van Gogh. The image of Seoul had a bit of a mix; half of it was traditional hanok and the other half was modern buildings. However, the hanok didn't really look like a hanok and seemed a bit strange, reminiscent of Japanese-style used goods. So, I felt a bit disappointed about that, but there were still some good aspects to it.

I asked for a sushi combo plate to be drawn, requesting eight pieces, including golden fried fish. However, it seems a bit different from what I had in mind. Nonetheless, when I look at the generated image, it reflects what we typically think of as sushi, which includes sashimi. There are also pieces like salmon sushi and tamago (egg sushi), along with some tempura. The drawing is quite satisfying and looks realistic. Next, I requested a fruit basket to be drawn, and it showed me a beautifully rendered image with depth and dimension. They usually present four images at once, and among them, this one turned out quite well.

I have only sketched what has come out. In conclusion, I briefly explained Microsoft’s Bing Image Creator. Next time, I will try out Google’s Bard, and I will provide an introduction to it. Please like and subscribe. Thank you.

마이크로소포트 빙 이미지 크리에이터(Microsoft Bing Image Creator) 소개 및 사용법  feat. by 구글 바드 (Google Bard) 일반 공개

국내에서 가장 많은 사물인터넷 관련 서적을 출간하고 삼성전자, LG전자, 한샘 등의 스마트홈 전략을 자문한 김학용 박사(전 순천향대 교수)의 유튜브 채널입니다. 이 채널에서는 개별적인 디바이스 소개보다는 사물인터넷(IoT), 지능형 사물인터넷(AIoT), 디지털 전환, 스마트홈, 아마존, 앰비언트, 캄테크, CES 등 최신 ICT 트렌드 및 비즈니스 전략을 소개합니다. 

출연 및 비즈니스 문의: iotstlabs@gmail.com



김학용 IoTV

마이크로소프트의 빙 이미지 크리에이터와 구글의 바드

Microsoft는 텍스트 입력으로 이미지를 생성하는 AI인 빙 이미지 크리에이터를 출시하여 텍스트 기반 응답에서 벗어나고 있습니다.

구글도 생성 AI인 바드를 실험적으로 공개했지만, 현재 미국과 영국에서만 사용할 수 있습니다.

Bing 이미지 크리에이터는 모든 브라우저를 통해 액세스할 수 있으며 이미지 설명을 위한 프롬프트 입력 상자가 있지만 영어만 지원합니다.

Microsoft ID로 로그인하면 하루에 최대 25개의 이미지를 생성할 수 있으며, 이미 17개의 이미지가 테스트되었습니다.

아이디어 탐색 섹션에는 이전에 만든 작품의 예가 제공되며, 작품 섹션에는 생성한 이미지가 나열됩니다.

주목할 만한 예로는 몇 달 전에 만든 빛나는 구슬을 가진 귀여운 햄스터와 우주비행사 시바견이 있습니다.

남자의 그림은 잘 그려졌지만 손가락이 네 개이고 담배가 떠 있는 것처럼 보입니다.

여자의 시선은 이상하게도 한 눈은 오른쪽을, 다른 눈은 왼쪽을 바라보며 긴장된 표정을 짓고 있습니다.

피카소, 일본 애니메이션, 픽셀 아트 등 다양한 스타일을 요청해 독특한 해석이 나왔고, 반 고흐 스타일의 서울 그림은 전통과 현대적 요소가 섞여 있지만 한옥이 진짜처럼 보이지는 않았습니다.

초밥 콤보 플레이트는 황금빛 생선 튀김, 사시미, 연어 초밥, 타마고, 튀김 등 총 8가지로 구성되어 만족스럽고 사실적인 이미지가 완성되었습니다.

생성된 과일 바구니 이미지는 깊이감과 입체감을 살려 아름답게 렌더링되어 제시된 4개의 이미지 중 단연 돋보였습니다.

필자는 다음에는 구글의 바드를 사용해 볼 계획이며 이에 대한 소개를 제공할 예정입니다.