MoltenArt V2 (SD 1.5): v1.0: vae-ft-mse-840000-ema-pruned.safetensors

Yet another SD 1.5 Checkpoint merge…New and Improved! (Please see version notes)

This is a Checkpoint merge model that I’ve been working on for a little while with the goal of achieving a different (even slightly) look to the results, especially in the face, from other models out there so far. I’ve tested this model against other popular models, and I prefer the results of this model most of the time.

I am sharing it in case there are other people that like its results also.

This is a realistic model that can handle fantasy, Loras, lycoris, textual inversions, Booru tags, simple or complex prompts, etc.

The Negative Prompt can have a large affect on the results of any model, so if the results are not to your liking please try a different negative prompt. I have posted many images that I have generated with this model, so feel free to copy the negative prompts I've used when trying out this model. (Please make sure to have the correct Embeddings installed!)

Sample Negative Prompt: "ng_deepnegative_v1_75t, (badhandv4:1.2), (worst quality:2), (low quality:2), (normal quality:2), lowres, bad anatomy, bad hands, ((monochrome)), ((grayscale))"

Another Negative Prompt: "bad-hands-5, (worst quality:2), (low quality:2), lowres, bad anatomy, ((monochrome)), easynegative, badhandv4, watermark, signature,"

Settings used for the examples:

VAE: vae-ft-mse-840000-ema-pruned ( )

base resolution: 576 x 768

Euler A or DPM++ SDE Karras

Sampling Steps: 23-28;

CFG Scale 7-8

Hires.fix with 20 Hires Steps using 4x-UltraSharp_Balanced upscaler

Upscale by 1.6

Denoising strength 0.4 – 0.5

Clip Skip 2

ENSD “31337”

After Detailer ( can improve faces in the distance

Brightness Adjustment_B_1.1|밝기 조정|明度調整 Lora ( )to lighten images that are too dark.