Nowości w edytorze obrazów Midjourney &mdash; rewolucja w edycji (film, 13m)

In this video, we will take a detailed look at the MidJourney's new Image Editor update. New Image Editor allows us to edit external images which are not generated by MidJourney, meaning we can in-paint, resize, and even crop external images via a new editor. Additionally, we have a new feature called Retexture. When you upload an external image, it scans the underlying structure, shapes, and proportions of the original scene. You can then apply new styles, colors, patterns, and details based on your prompt while ensuring that the initial scene remains intact. This feature is currently open for a limited amount of users. In order to access it, you need to have at least 10,000 images generated. It's also open for people with yearly memberships and people who have been monthly subscribers for the past 12 months. Here you can see one particular example where they uploaded an external image and they added a flying whale to the scene. Here's another example from a real scene in US. They uploaded this external scene and added this metallic circle here. Let's test this out together. I'm clicking on Next and there is Terms of Service. I'm accepting it and clicking on Start Creating. This new Image Editor is placed on the left-hand side under Edit button. There are two options. You can upload an image from your computer or you can upload from a URL. I will upload the photo of my dog Lonnie and we will edit it a little bit. It's an external image. It's not generated by McJourney. You have two sections in the editor. One is Edit and second one is Retexture. So it says Retexture will change the contents of the input image while trying to preserve the original structure. We will test this out, but first let's start with Edit. This is the usual McJourney Image Editor that we tested before, but there are some new things here. First of all, you can crop images now. I clicked on Move and Resize. I pulled the handle here and I think I still need to add a prompt and I clicked on Download Image, but it didn't download it. I think we need to first reimagine it. I click on Submit Edit. It says Starting. So Image Generation, User Interface also looks a little bit different here. Now we also see percentages. Okay, now it's cropped, basically, but I needed to write the prompt again. For future, probably for cropping, they can just allow people to use these handles and download the image without trying to reimagine it. This is probably one thing to improve in the future because I was a little bit confused. I thought I can just crop the image, but I understand now that you need to reimagine first. Next thing is we can erase elements in our image. So I can change price size and I will get rid of this back here because it's not necessary in our Scene. And I will just write the same prompt and I will hit Edit. So this should now erase this back from the Scene. It is reimagining. Okay, and it's gone. So there are also a few other options. You can choose the one who looks most natural. I think first option is the best. Now let's have fun a little bit. Okay, now I'm going to choose this part and give Lonnie some cool sunglasses. A dog with cool sunglasses sitting on a sofa. I will also add my personal style with dash dash P, my personalization code, which tends to be much more realistic. It's not really illustrative. So let's hit Submit Edit. And we are expecting now to see sunglasses that Lonnie is wearing. It has this cool animation of kind of like it's changing color. Okay, that's pretty awesome. Now Lonnie has some sunglasses. Pretty cool and it's a realistic looking one as expected. Now let's give Lonnie a T-shirt. I selected the upper part and I wrote a dog wearing a T-shirt with text, I love poop. She has cool sunglasses and she's sitting on a sofa. Let's Submit Edit. All right, T-shirt is here. The text is off, unfortunately. So text didn't quite work out well, but T-shirt is looking realistic and perfectly nicely adjusted. Only the text is problem. Maybe we can fix that. So if I choose here. All right, so I choose the text and I use the exact same prompt, but I couldn't get the text that I was hoping. Now, as always, we can change the aspect ratio. So let's make this one-to-one. So we can actually test the Expand feature as well. So I will just like resubmit the same prompt and I'm expecting that we will have a cool expansion of the same scene here. All right, so we got the expansion and there are a few options that we can choose. This one looks the best for me. And with this image, you can do multiple things. You can upscale this image to your gallery. So it will upscale this image. The full quality version of your image will be added to your gallery or you can just simply download it and use an external upscaler. I'm going to download this. It's done. And let's try retexturing. I added steampunk style photo of a dog to the beginning of my prompt. It added this R icon, meaning retextured. All right, so now we have the results. So it made it with the steampunk colors, almost like a photo filter. Maybe this looks really steampunky. And let's also try... So I wrote dog wearing a T-shirt with text. I now put with complete steampunk outfit. She has cool steampunk sunglasses and she's sitting on this. Let's update this one and see what happens. All right, so we have our images. We got kind of the desert scene and steampunk glasses. Okay, so this looks pretty cool. And this looks a little bit strange, but it's really steampunky. And it also changes the color of Loni. The same prompt, I just specified the color of Loni. That's cool. That's much better. The prompt box is extremely sensitive, so you really need to specify exactly what you need. All right, this one is pretty funny. So I will download this. Okay, and if you click on Edit again, you can see where you left. So you can come back to where you left directly with a single click. And you can see all edits you generated. And of course, obviously, you can start editing a new image. Next use case I want to show you is product photography. The first time in Midjourney, you can upload your own logo or custom illustrations and create product photography with it. So let's upload my Cyber Jungle logo. And I'm going to expand it to one-to-one, make it a little smaller, and place the logo to the center. And I'm going to write, print it on a black mark, and I will submit edit. It didn't work out quite well, as I was expecting that it would place the logo on the mark. And I realized while reimagining, it changes the logo quite a bit. So this didn't work out. I also tried product photos. Cyber Jungle logo is printed on a black mark, ready for e-commerce. Let's submit this one. Okay, guys, unfortunately, this use case is not really working. Or maybe I couldn't figure it out. But that would be actually a great use case because many people were always asking to upload their custom logo to their product photos in Midjourney. And this is a huge use case. Let me know down in the comments if I missed something. And hopefully, they will add this feature because this would be really fantastic. Okay, next use case is around architecture. So I will upload this image of a wooden chair in an empty room. We have now also suggest prompt option, which is using the underlying describe engine that they have. And when I click on it, it will automatically populate the prompt box based on this image. A wooden chair in the corner of an empty room with the sunlight shining through the window, blah, blah. This saves time, and I think it's a really good idea that they added this here. It's just maybe suggest prompt should be somewhere closer to prompt box. Probably UX can be improved here. But nevertheless, it's a good feature. I'm going to choose this wooden chair now and update my prompt. A shiny metallic chair in the corner of an empty room. I hit submit. So simply chair looks same. The same size, shape, perspective, everything same, just texture changed. Now it looks more metallic. I'll try something else. So I want to have a green furry cushion here. This one looks quite furry. I will stick with this one. And now let's use retexture and try to make this scene a night scene instead of a day. So I updated my prompt a little bit. I added moonlight shining through the window in a night scene. So overall, it's a dark environment, and I removed the bright environment from my prompt. Let's submit retexture and see what's going to happen. Okay, interestingly, it's got definitely darker. It's a darker scene. I'll say maybe this comes close. This looks like also an evening scene. And now we have the furry green cushion much more visible. You can see, especially in this one, pretty good. So let's make it not an empty room, but of a jungle room with full of wild plants. Okay, interesting. It put the jungle outside. Maybe we are conflicting with minimalist space. Okay, I mean, it's definitely not Midjourney's fault because there were some extra parts where I didn't realize it was conflicting. It came close to what I was imagining. Now let's go crazy and make everything latex. I will just say a latex chair in a room made of latex. So everything's latex. Okay, we have the latex chair. It's there. The room is okay. Maybe this one also looks latex. It's pretty fun. Look at this. So we can also add a style reference here. You don't have to only add your personalization code. You can also add the style reference. I have this style reference code I saw online. It's pretty illustrative and it has nice colors. So I will submit this with the style reference. Wow, look at that. Yes, made it more illustrative. It's less realistic now. Put the style reference from the code I entered. Pretty impressive. Another test. I'm going to ask it to suggest prompt because I don't want to lose time by writing. Louis Singh, cinematic shot of Kate Winslet. Okay, is an exotic pirate captain. I added, it's nighttime with visible stars. I switched to retexture and I added, it's nighttime with visible stars. Let's run this and see what happens. Okay, changed scene to night definitely, but it also changed our character. To prevent that, we probably need to use also character reference. I uploaded the same image here and I will choose this as a character reference and submit retexture again and see that if it is going to keep the character more consistent. Hmm, well, I would say we still lose a little bit of the original character's look. These are all different characters, so it's difficult to say that the character reference was quite effective. What would happen if I would submit this as a regular edit? We go back to this image and choose the face and I have my character reference. I submit it. Oh no, well, this looks not that great. Okay, we lost the original character's face. Well, this was the first test and hopefully this video was truly helpful for you. We will continue playing with this image editor and explore further. And don't forget to give this video a thumbs up if you found it helpful and make sure to subscribe for more in-depth tutorials. If you want to learn more about AI creativity, click here.

Menu

Nowości w edytorze obrazów Midjourney — rewolucja w edycji (film, 13m)

Toggle timeline summary

Transcription

Menu