Hi everybody, happy new year! This was my last model update right before I hit the holidays, and is my first public release of 2024 and I'm excited it's DynaVisionXL! This time around I've performed 5 trainings targeting improving widescreen aspect ratio output, improvements to eye and pupil quality, These trainings I used my own training tools and methods that have shown to improve coherence and prompt accuracy, even for more complex prompt situations.
Changelog 1/13/24
5 total internal tunings applied
3 trained on varying wide and tall images
2 focused on coherence and detail, especially eyes
low-value back-merge with DV534 to help stabilize the DynaVision style
Known issues
male genitalia
hands
medium distance mushiness for faces and fine details (this is a VAE issue, but still worth tracking here)
DynaVisionXL_v0.6.1.0 Release Notes
Hi everybody, happy new year! This was my last model update right before I hit the holidays, and is my first public release of 2024 and I'm excited it's DynaVisionXL! This time around I've performed 5 trainings targeting improving widescreen aspect ratio output, improvements to eye and pupil quality, These trainings I used my own training tools and methods that have shown to improve coherence and prompt accuracy, even for more complex prompt situations.
Changelog 1/13/24
5 total internal tunings applied
3 trained on varying wide and tall images
2 focused on coherence and detail, especially eyes
low-value back-merge with DV534 to help stabilize the DynaVision style
Known issues
male genitalia
hands
medium distance mushiness for faces and fine details (this is a VAE issue, but still worth tracking here)
long necks in tall aspect ratios