MOSS-Image-SFT

Quantised media packages of OpenMOSS-Team/MOSS-Image-SFT.

MECHA Forge

MECHA — Model Engineering & Contextual Hearth Assembly

These media packages have been forged and optimized in the MECHA Forge. Each one holds the Denoiser, Text Encoder, and VAE components packaged together for easier deployment.

Description

MOSS-Image-SFT: image generation model with multi-language prompts, layout control, and text rendering. Base model, undistilled.

Available quantizations

Quant	Denoiser	Text encoder	VAE
FP4	FP4	FP4	FP8
FP8	FP8	FP8	FP8
INT4	INT4	INT4	INT8
INT8	INT8	INT8	INT8

The VAE is shared between formats: FP4 reuses the FP8 file, INT4 reuses the INT8 file.

Recommended settings

From the source model card (adjust to taste):

Parameter	Value
Steps	50
CFG / Guidance scale	3.0 (range 3.0-6.0)
Resolution	1M-2M total pixels
Notes	Shift 5.0, Euler sampler, negative prompt empty

Example prompts

Caught in the moment (candid, slice of life):

masterpiece, best quality, highres, absurdres, very aesthetic, safe, 1man, solo,
candid moment, natural lighting. A cozy bookshop interior in the afternoon,
wooden shelves, dust motes in sunlight, a reading nook by the window. A 30-year-
old man with tousled dark hair and glasses. He has a lean build and a thoughtful
expression. He wears a knit sweater and chinos, holding an open book, mid-smile,
candid and relaxed.

Fashion photography (magazine cover style):

Fashion photography, magazine cover style: a 25-year-old man with tousled dark
hair, strong brow, and stubble; wearing a tailored charcoal overcoat over a
turtleneck, leather gloves, Chelsea boots; moody backlight, editorial layout,
high contrast, 8K, sharp focus.

Scenery:

Scenery: a misty bamboo forest at dawn, soft green light filtering through tall
stalks, a stone path, dew on leaves, tranquil atmosphere, photorealistic, 8K.