This adds some of my "sanity" modules which seem to have made the model generate bad hands and extra limbs a bit less - but the negative hands embedding(s) are still useful in many situations. (I really don't think this is a problem that can be absolutely solved in 1.5 models - but we can make things better).
The biggest change is a (fairly light, "Trained Difference Only") merge of Part_LoRA's Neutrino Trained Model. As described on that model - the output is unrefined and actually needs a merge to get some output consistency. Normally I run away from things like that - but in testing it, it really does do some fun things that other models don't. With the merge, those concepts seem to be working pretty well - and it didn't negatively affect the look and feel of this model so... it's now canon.
The new triggers aren't outlined all that well on the Neutrino page, but they do give you some good clues for things to try. Because it's in there fairly lightly, you may need to (emphasize) the tokens to get them really showing, right. (emphasis:1.2) or 1.4 seem to be enough in most cases.
This adds some of my "sanity" modules which seem to have made the model generate bad hands and extra limbs a bit less - but the negative hands embedding(s) are still useful in many situations. (I really don't think this is a problem that can be absolutely solved in 1.5 models - but we can make things better).
The biggest change is a (fairly light, "Trained Difference Only") merge of Part_LoRA's Neutrino Trained Model. As described on that model - the output is unrefined and actually needs a merge to get some output consistency. Normally I run away from things like that - but in testing it, it really does do some fun things that other models don't. With the merge, those concepts seem to be working pretty well - and it didn't negatively affect the look and feel of this model so... it's now canon.
The new triggers aren't outlined all that well on the Neutrino page, but they do give you some good clues for things to try. Because it's in there fairly lightly, you may need to (emphasize) the tokens to get them really showing, right. (emphasis:1.2) or 1.4 seem to be enough in most cases.