Features are nice, but the generation feels of worse quality and is now completely lobotomized. Feels like the main purpose is to be a quota eating machine at which this model actually succeeds beyond all expectations:
- Anomalies
- Complex character poses need much more attempts than before (multiple legs, hands, ignored prompt, etc.)
- Less face variation
- Changes to still images feel even more as if poorly photoshopped in than before
- Reused prompts result in worse quality compared to prev versions
- With seemingly higher photorealism uncanny valley hits harder than ever
@takatukiaran You can place a square roof piece instead. It usually works in any such situation and in your case with all not be noticeable since it's in the middle of foundation pieces