This (and the other post by this artist) look like they might have been at least partially made with AI. Notice that in the other work that has been uploaded by this artist, Asuna is wearing very similar clothing, but not exactly the same-- the shirt fits more tightly, there's a boob window, and the booty shorts are jean shorts instead of exercise shorts, and she has a choker in one image but not the other. In short, these two Asunas look like they come from basically the same prompt, or a very similar one. This is a common problem with the AI-- getting an outfit to be consistent across multiple images, even with an identical prompt, is nearly impossible, but the outfit will always be the same in "broad strokes" (long sleeve white shirt showing midriff, blue short shorts, blue hair ribbon). Further, this is one of the reasons AI pictures of popular characters are so often in alternate costumes-- because the costume is likely to generate with incorrect details, so the person generating the image either has to accept this dead giveaway that they are using AI or put in extra time correcting those details. So, it seems likely to me that the "artist" generated these images of Asuna, probably on a white background, the cropped and copy/pasted them onto the backgrounds we see in the final image, with the bubble effect as a compositing trick which makes it harder to notice any seams where the crop isn't perfect. It's also possible that the images were generated whole-cloth, but that seems less likely to me. At the very least, then, it seems these images are AI-assisted at best, although it always leaves a sour taste in my mouth when people aren't honest about using AI, and when the work that they put into actually editing those images is primarily to facilitate that dishonestly. As such, I'm going to add the AI-generated tag, but I do acknowledge that it seems likely that some amount of manual editing did happen to this image. I just don't think it adds enough to make it worth keeping on danbooru.