 So you generated some cool stuff with your character and you've got some promising images but the quality just isn't good enough for you to use in your retraining. Not a problem. If you find an image that you really like but it needs some fixing, there are multiple ways you could address this. If you are an artist, sometimes it's faster to just manually clean it up yourself in Photoshop. But in case you're not, maybe you're a programmer, maybe you're an engineer or a doctor or something else very important. And if that's the case, there's a few easy tricks you can do right here that will drastically fix the quality of the images you're not happy with. First click the recycle icon to keep the seat of your image the same. Then go to the top left and hit high res fix and click the arrow next to it and open the full menu. Change upscaler to this one here D noise strength to 0.4 and what this does is turn your ugly fuzzy 512 by 512 image into a much less fuzzy 1024 by 1024 image. The reason why we upscale this way instead of just hitting high res is because if you just hit high res, you're basically just upscaling a compressed blurry image. So use these other settings when you do this. If we generate now you can see the quality of the images is already looking much better. So now if we find an image that we really like, we can click it, click here to send it to image to image, right click on the image, copy, go to in paint, control V to paste. And now you can mask any of the areas you're not happy with. Let's just say I wanted to fix her face. Just left click to mask that area out. If you mess up, you can always click the undo button here and we only want to try and regenerate the masked area. So we're going to check only mask and padding controls how much the mask will blend with the areas bordering your mask, the higher this number, the bigger the area of blend between the masked area and the unmasked areas. Now it defaulted to 1024 because we did some high res stuff, but stable diffusion 1.5 generally works with 512. So we'll just use that instead. Okay, now we're ready to generate the selected areas again, but we should change the prompt a bit before we do this because we are just fixing the face. We don't want to describe the entire body. So other things like the hair, ponytail and shorts, we can remove those. We just want to generate the face and look at that. Her face has basically been fixed and check this out. Let's say you're working on a manga or a webtoon or an anime and you just need to change the facial expression of your character. Well with your trained Laura, all you got to do is literally type the emotion or the expression you want. Let's just say I wanted blended a cry, type it in here and now we have her face crying. And last, if you get some crazy looking stuff that usually means that your denoising strength is a little too high or the padding pixels are too low. So you might want to try reducing the denoise strength to something like 0.65. So mess with those if you're getting some cursed looking results. Anyways, hope that helps and as always hope you have a fantastic day and I'll see you around.