This is a simple Text-to-Image workflow for the Moss model (dirtyMoss).

It features a two-stage generation process:

1. Initial generation at 768×1344

2. Light refine pass (denoise 0.35)

3. Upscale with remacri

Generation speed: ~35–41 seconds per image on RTX 4060 Ti 16GB.

Required models:

• UNET: dirtyMoss_int8.safetensors

• CLIP: Qwen3.5-9B-heretic-v2-merged-int8-convrot.safetensors (moss_image type)

• VAE: flux2-vae.safetensors

• Upscaler: remacri_original.safetensors

LoRAs used (via Power Lora Loader):

• cndamiph_Candid_Amateur_iPhone_MOSS_v1 (0.2)

• 8-steps-dmd2-lora (1.0)

• realistic-moss-sft-nikastyle-moss-image-sft-epoch-15 (0.8)

• PornMaster_Moss_Detail_Slider_V1 (4.0)

Custom nodes:

• Power Lora Loader (rgthree)

Because I'm very lazy, I didn't make it pretty and didn't group the nodes — it'll do as is. The workflow is very simple and straightforward.

Just drop your prompt, queue, and enjoy. Works well for photorealistic portraits and full-body shots.