Trained on 1024 resolution. Requires more vram for run than 768, but gives much more detailed results.
This version works better in some cases and worse in some cases. As for me, this version is more stable and works better with non-structures.
Important: This model is trained on 1024 resolution pictures so you must use at least 1024 dimensions when run. But this works perfectly, if you run more than 1024 (e.g. 1024x2048) generating just bigger battlemap.
Important #2: V 2.0 Do not require any trigger words. Just describe normally.
Trained on 1024 resolution. Requires more vram for run than 768, but gives much more detailed results.
This version works better in some cases and worse in some cases. As for me, this version is more stable and works better with non-structures.
Important: This model is trained on 1024 resolution pictures so you must use at least 1024 dimensions when run. But this works perfectly, if you run more than 1024 (e.g. 1024x2048) generating just bigger battlemap.
Important #2: V 2.0 Do not require any trigger words. Just describe normally.