Tensor.Art
Create

MOSS IMAGE - MOOS BASE REALISM

MOSS IMAGE

CHECKPOINT
Original

Updated:

MOOS IMAGE SFT
MOOS SFT REALISM UP
MOOS BASE REALISM
MOOS SFT REALISM
MOSS IMAGE by REKTY ANJANY on Tensor.Art
MOSS IMAGE by REKTY ANJANY on Tensor.Art
MOSS IMAGE by REKTY ANJANY on Tensor.Art
MOSS IMAGE by REKTY ANJANY on Tensor.Art

MOSS Realistic — REKTY Edition

A photorealistic checkpoint based on MOSS-Image SFT (5B DiT by OpenMOSS), developed and tuned in-house by REKTY. Built for high-quality realistic portraits: natural skin texture, detailed fabric weave, strand-level hair, and clean color rendering — while keeping MOSS's strength in rendering text inside images (signage, posters, screens).

just load it, no extras needed.

Recommended settings

ParameterValueSamplerer_sdeSchedulersimpleSteps25CFG2.5Resolution832×1216 (portrait) / 1216×832 (landscape)Negative promptempty

Please use the er_sde sampler with the settings above for best results. Other samplers may produce a softer or hazier look.

Prompting tip

This model performs best with full scene captions — one paragraph containing: subject + outfit + location + lighting + camera/lens + color palette + mood + style.

Example:

Editorial fashion portrait of a young woman with long black hair, standing in a narrow alley at dusk after rain, oversized charcoal wool coat over cream turtleneck, wet asphalt reflecting sodium streetlights, shot on 85mm f/1.8, soft key light from a shopfront, muted teal and amber palette, fine film grain, calm mood.

Short 3–5 word prompts will produce generic output far below this model's capability.

Best for

  • Portrait, editorial, fashion, street photography

  • Realistic Asian and global characters

  • Scenes with signage or text in frame — MOSS renders text natively

Version Detail

MOSS-Image-Base
4000
10
# MOSS Image BASE — REKTY Edition A portrait and BEST-photography checkpoint based on **MOSS-Image-Base** (5B DiT by OpenMOSS), developed and tuned in-house by **REKTY**. Built for polished close portraits: clean skin texture, strand-level hair, soft directional lighting, restrained backgrounds, and editorial color rendering — while keeping MOSS's strength in rendering text inside images (signage, posters, screens). Everything is a single file — just load it, no extras needed. **Start your prompt with the trigger word:** `moss_UR IDE_style,` ## Recommended settings | Parameter | Value | |---|---| | Trigger | `moss__style,` at the start of the prompt | | Sampler | **er_sde** | | Scheduler | **simple** | | Steps | **30** | | CFG | **3.0** | | Resolution | 832×1216 (portrait) / 1216×832 (landscape) / 1024×1024 | | Negative prompt | empty | > Please use the er_sde sampler with the settings above for best results. Other samplers may produce a softer or hazier look. ## Prompting tip This model performs best with full scene captions — one paragraph containing: **subject + outfit + location + lighting + camera/lens + color palette + mood + style**. Short 3–5 word prompts will produce generic output far below this model's capability. **Best for** - Beauty close-ups, editorial and fashion portraits - Natural skin texture and detailed hair - Realistic Asian and global characters - Scenes with signage or text in frame — MOSS renders text natively ## Licensing & attribution Based on MOSS-Image-Base by OpenMOSS (https://huggingface.co/OpenMOSS-Team/MOSS-Image-Base), modified and distributed under the Apache License 2.0. Model modified by REKTY. Not endorsed by OpenMOSS.

Project Permissions

Reprinting is strictly prohibited

    Use Permissions

  • Use in TENSOR Online

  • As a online training base model on TENSOR

  • Use without crediting me

  • Share merges of this model

  • Use different permissions on merges

    Commercial Use

  • Sell generated contents

  • Use on generation services

  • Sell this model or merges

Related Posts