Model Card for MossyMix v2

This is a merge of a bunch of MOSS Image LoRAs into MOSS Image Base (not SFT) The focus was mainly on photorealism and optimizing for output that emulates smartphone imagery or older media like analog film, but it can also do some other cool stuff.

Model Details

Model Description

• Model type: Checkpoint
• Finetuned from model: MOSS Image Base

Uses

Don't use this for bad things.

How to Get Started with the Model

Run these models in DrawThings with the following settings (ComfyUI needs slightly different ones also listed here):

MossyMix v2

• Download here
• Resolution: ~1.5 MP preferred (for example 1536x1024 or 1728x1152)
• Text Guidance / CFG: 3.0
• Sampler: DPM++ 2M Trailing
• Steps: 30
• Shift: 5.00 or 4.00
• In ComfyUI use res_multistep / linear_quadratic or euler / simple, CFG 3.5, 30 steps
• If you want to run this with CFG 1 at 15-20 steps, add 0.55 (55%) of DMD2 LoRA

MossyMix Turbo v2 (includes DMD2 LoRA)

• Download here
• Resolution: ~1.5 MP preferred (for example 1536x1024 or 1728x1152)
• Text Guidance / CFG: 1.0
• Sampler: DPM++ 2M Trailing
• Steps: 15+
• Shift: 5.00 or 4.00
• In ComfyUI use res_multistep / simple, CFG 1.0, 15+ steps