One more time about the DLSS and unnecessary hysteria around it. This is Katherine from KCD. She is the well-known Czech actress Kristýna Leichtová. The first image shows a photo of her. This is how we 3D-scanned her. We then manually adjusted her appearance and textures, because we don’t use any makeup during the 3D scan, and the in-game model has to meet a whole range of parameters. The second image shows her appearance in the vanilla game, the third shows DLSS5 when it’s set correctly (in my opinion), and the last one shows what happens when random idiot sets all the sliders to max. We can discuss whether you prefer the slightly “smoothed-out” original look from image 2 or the more natural one from image 3; in any case, in image 3, the in-game model is clearly the most similar to the real actress who plays Katherine and whom we scanned. It’s also much better lit, and the details of the nose, eyes, hair texture, and subtle shadows, or the bones on the neck definitely look more natural and realistic, and DLSS5 isn’t adding anything artificial there. Those details are present in both the model and the texture; it’s just that due to the engine’s lighting, they aren’t as visible, and DLSS5 “fixes” this fairly accurately based on the normal map and other factors. On top of that, this particular technology will not steal anyones job. So, for me, Image 3 is an improvement and much closer to how I’d wanted the lighting to work from the start and it was not possible. If we had this tech two years ago, and I would like the character to look "younger" but with DLSS5 shadows and lighting, we would tweak the model and made it so. Of course, this raises the issue of the uncanny valley. It sometimes starts to look almost too realistic—similar to the “soap opera” effect in movies when you turn on frame interpolation on a TV. As creators, we’ll have to deal with that. But it’s a new technology, and I’m absolutely certain it will improve.

GPT Image 2.5 vs. Nano Banana Pro on Higgsfield. The challenge: consistent 360° character and room sequences from just one reference image per scene.

GPT 6 Astra is INSANE 🤯 It is a BEAST in 3D MODELING!! This is our first test, it blew our mind 🤯 It was able to rebuild a building infront of our house from just a few iphone photos in Blender in less than 30 minutes …

Yes this was just one prompt! Image reference and it made that in Blender, you can't understand my shock. It can make anything in Blender. My prompt was super basic, literally "Make this in Blender" with the second image attached lol

🔥 A wildfire W of Athens 🇬🇷 has burned over 10k hectares, triggering evacuations & a large-scale emergency response. 🛰️ This #CopernicusEU Sentinel-2 #ImageOfTheDay captures the burn scar and a 70 km-long smoke plume. 🔗 eu-space.europa.eu/explore-e…