MOSS-Image is a 5-billion-parameter open-source text-to-image model released by OpenMOSS in October 2026. It uses a Diffusion Transformer (DiT) architecture and is released in two variants: Base (for LoRA fine-tuning) and SFT (for general-purpose generation — that's what runs on this site).

Is MOSS-Image free to use?

Yes — this site provides a hosted version of MOSS-Image that starts with free credits. No HuggingFace account, no GPU, and no installation is required. Paid plans are available if you need higher volume.

What's the difference between MOSS-Image Base and SFT?

MOSS-Image-Base is the pretrained foundation model intended for LoRA fine-tuning and research. MOSS-Image-SFT is the supervised fine-tuned version optimized for general image generation — that's what this site runs.

Can MOSS-Image render text inside images?

Yes — text rendering is one of MOSS-Image's advertised strengths. It handles English text inside signs, posters, and graphics better than most open-source alternatives. Complex multi-line layouts may still need prompt iteration.

Do I need a GPU or ComfyUI to run MOSS-Image?

No. This site hosts MOSS-Image for you — everything runs in your browser. The official local setup requires ~11GB VRAM (FP16) and a ComfyUI or Python environment, which we eliminate entirely.

What styles does MOSS-Image support?

The official release highlights photorealistic and anime-style generation. Specify the style in your prompt (e.g. 'photorealistic portrait' or 'anime illustration') and the model will follow.

How is this different from the official HuggingFace demo?

The HuggingFace Space is free but requires a HuggingFace login and runs on Zero GPU with queue times and cold starts. This site hosts MOSS-Image on dedicated infrastructure — no login required to start, no queue, faster generation.

Can I use the generated images commercially?

MOSS-Image is released under the Apache 2.0 license, which permits commercial use. Images you generate are yours — check the terms for details on acceptable use.

Run MOSS-Image Online — Free, No Setup

OpenMOSS's 5B text-to-image model. Photorealistic and anime styles, clean in-image text, no GPU or HuggingFace account required.

0 / 2000

Try:

No HuggingFace loginPhotorealistic & AnimeIn-image text

What is MOSS-Image?

MOSS-Image is a 5-billion-parameter open-source text-to-image model released by OpenMOSS in October 2026. It uses a Diffusion Transformer (DiT) architecture and is explicitly trained for both photorealistic and anime-style generation — two styles most open models struggle to do well simultaneously.

5B DiT model, Apache 2.0 license — free for commercial use
This site hosts MOSS-Image-SFT, the supervised fine-tuned variant for general-purpose generation
Photorealistic and anime styles, in-image text rendering, long structured prompts

Learn more: OpenMOSS/MOSS-Image · Official HF Demo · API docs

Why Creators Pick MOSS-Image

What this 5B model actually does well — verified against the official release

📸

Photorealistic & Anime

MOSS-Image was trained for both photorealistic scenes and anime-style illustration — two of the most demanded styles in one model.

🔤

Clean In-Image Text

Renders legible English text inside posters, signs, labels, and graphics — a known weak point of many open image models.

📝

Long Prompt Support

Describe complex scenes in detail. Long, structured prompts are handled natively — no truncation, no keyword cramming.

☁️

No GPU, No Setup

Skip the ~11GB VRAM requirement and ComfyUI installation. Generate directly in your browser on our hosted infrastructure.

📐

Multiple Aspect Ratios

Square, portrait, landscape, and widescreen formats — pick the ratio that fits your use case before generating.

🎁

Free to Start

Apache 2.0 open model. We give you free starter credits — no HuggingFace account, no queue, no cold-start waiting.

How to Use MOSS-Image

From prompt to image in three steps

Write a prompt

Describe the scene in plain English — the more detail, the better. Long prompts are supported.

Pick an aspect ratio

Square for posts, portrait for covers, widescreen for banners — choose before generating.

Generate & download

Get your image in seconds. Download it or iterate on the prompt to refine.

✨ Generated with MOSS-Image

See what the model can do

Photorealistic

Photorealistic cinematic shot of a Japanese tea house at the edge of a bamboo forest in light rain, warm interior glow, 50mm lens, film grain, overcast rainy afternoon

Anime

Anime illustration of a teenage girl with short silver hair and amber eyes wearing an oversized olive-green hoodie, soft lighting, Kyoto Animation style, pastel background

Text in Image

Minimalist specialty coffee bag label, bold sans-serif text 'MOSS ROAST' on top line and 'SINGLE ORIGIN' on second line, small geometric mountain icon, moss-green background, flat modern branding

Concept Art

Photorealistic close-up of hands kneading bread dough on a floured wooden countertop, flour dust in the air, warm kitchen window light, documentary food photography

Photorealistic

Surreal floating garden above a calm lake at twilight, oversized glowing flowers and moss-covered stones levitating, dreamy soft-focus pastel look, Studio Ghibli meets modern 3D

Anime

Vintage WPA national park poster of a misty pine forest trail, text 'WANDER THE MOSS' at top in uppercase serif, screen-print aesthetic, sage green and cream palette

Text in Image

Cozy anime illustration of an orange tabby cat sleeping on a pile of books in a sunlit library, warm painterly Ghibli-inspired style, soft sunbeams through window

Concept Art

Photorealistic overhead flat-lay of a ceramic matcha tea set on linen cloth — matcha bowl, bamboo whisk, tea scoop, small mochi, soft diffused daylight, minimalist food photography

Try MOSS-Image Now

Simple Pricing

Start free, upgrade when you need more

Each generation uses a small credit amount. Free starter credits let you try MOSS-Image before committing.

Public Preview

Basic Plan

$29/month

2,900 Credits/month

Approx. 1,450 generations

✓Standard speed

✓Email support

✓MOSS-Image engine access

✓Guided prompt builder + presets

Recommended

Pro Plan

$59/month

5,900 Credits/month

Approx. 2,950 generations

✓Priority processing

✓Priority support

✓Batch background removal (beta)

✓Growing template library

✓Future access: MOSS-Image

Public Preview

Max Plan

$99/month

9,900 Credits/month

Approx. 4,950 generations

✓Highest priority

✓Dedicated support

✓Future multi-model compare workspace

✓Team workspaces & roles

✓Custom branding & watermark control

Need more flexibility?

View full pricing

Frequently Asked Questions

Everything you need to know about MOSS-Image

Start Generating with MOSS-Image

Free credits, no signup, no GPU needed.