Sign In

Deliberate Cyber

282
8.6k
108
Verified:
SafeTensor
Type
Checkpoint Merge
Stats
2,330
Reviews
Published
Oct 7, 2024
Base Model
SD 1.5
Hash
AutoV2
4ACE11B0AA
WillieF's Avatar
WillieF

Buy Me a Coffee: https://www.buymeacoffee.com/willief

Update:

Version 1 is a Merge of Deliberate V3 and CyberRealistic V4.1

Version 2 is a Merge of Deliberate V6 and CyberRealistic V4.2

Version 3 is a Merge of Deliberate V6 and CyberRealistic V5.0

Version 4 is a Merge of Deliberate V6 and CyberRealistic V4.2 and The Workshop V1.

Version 5 is a blend of Deliberate V6 and CyberRealistic V6, but I added The Workshop V1 to allow the Checkpoint to be more creative and vibrant. This is easily the best version, so far, thanks to Deliberate V6 and CyberRealistic V6 being the best versions of their respective Checkpoints.

Also, each version now has a pruned file (thanks to aueki4g467 for the direction). The full files can still be found in the Files area to the right.

Original Description:

This is a merge of Deliberate_V3 and CyberRealistic_V4.1.

So, why merge these two checkpoints?

To me, CyberRealistic is hands down the best SD1.5 Checkpoint I've used. It makes the most reliable pictures (without any deformities or other typical issues), it is the most accepting of just about any Lora I've ever tried, and it is pretty accurate when it comes to prompt recreation. If it lacked anything, it was flair. If you wanted to prompt every detail, CyberRealistic would most likely make it happen (within reason). However, if you didn't manage to write a descriptive prompt, it was going to essentially just do exactly what you told it.... which is great most of the time, don't get me wrong.

On the other hand, Deliberate_V3 is all about the flair and the subtle nuances (I know it's currently on V5, but V3 is still the best). With very little prompting, it can make a detailed and nuanced picture often taking creative liberties to fill any space you didn't prompt for yourself. What it is kind of bad at is integrating any Lora I've ever tried because it likes to take the same creative liberties. Also, even when things do come close to the Lora, everything always comes out looking too airbrushed for my tastes. So, why did I stick with it? Because it often output gorgeous pictures in ways I never even considered. It was a fun Checkpoint that had the potential to surprise you.

What I wanted was the best of both worlds. I wanted the reliability and flexibility of CyberRealistic, and the beauty and creativity of Deliberate. Through trial-and-error (and some light reading on Automatic1111 Checkpoint Mergers), I think I finally hit the sweet spot. It's everything CyberRealistic has always been, but with a little more flair and a little easier on the prompting.