4 ms·
Not a single human face in these samples. I wonder how well it does on faces.
by polotics 4y ago
Not a single human face in these samples. I wonder how well it does on faces.
- tmikaeld 4y agoI’m also curious. The animal faces and eyes look good, less creepy than most.
- 8n4vidtmkvmk 4y agohave you seen https://thispersondoesnotexist.com/ https://thispersondoesnotexist.com/ ? that's old already. different tech tho
- andresgottlieb 4y agoSaul Goodman if he was a character in Twin Peaks: https://media.discordapp.net/attachments/999426920376717513/1011642719208214528/agottlieb_Saul_Goodman_as_a_character_in_Twin_Peaks_HDR_4K_f931186a-233e-4a70-86c8-76a78098d43f.png https://media.discordapp.net/attachments/999426920376717513/... Generated with Midjourney Beta
- glenneroo 4y agoWhich is now Stable Diffusion under the hood, sprinkled with a little prompt parsing "magic".
- Terretta 4y agoThis may be misleading. As far as I could tell (using it before, during, and after this Beta option was available) it was the upscalar using that, not the original 4 image generation.
- ansible 4y agoThe samples I was shown definitely had a theme: robots, raccoons, hats, and raccoons wearing hats. Our brains are very thoroughly wired to detect faces in general, and flaws in faces. So we have a very, very high standard for what passes muster. We are apparently much more forgiving with regards to what raccoons and corgis look like.