AI image workflow
Type a prompt and generate with MOSS Image, the open-source OpenMOSS model built for long Chinese and English text inside images. Put exact words in quotation marks.
Open-source text-to-image model
Prompt examples
Current workflow
A result appears only after you run the workflow; no simulated AI output is shown here.
Model for your run
Text prompts only, no sample outputs. Pick one, then generate with the prompt, ratio, and model already filled in.
01 · Template direction
Event, travel, and promo posters with a short headline.
A vintage travel poster of a seaside town at sunset. Bold hand-lettered title at the top: "LISBON". Line at the bottom: "Sun, Sea & Tram 28". Flat colors, textured paper, balanced layout.
Proofread the title and bottom line before publishing.
02 · Template direction
Testing multi-line Chinese text inside a realistic scene.
A wide view of a refrigerator beside a kitchen counter with a fruit bowl. A small sticky note on the refrigerator door is handwritten in Chinese: "记得买牛奶和鸡蛋！牛奶两盒，鸡蛋一打。"
Small handwriting can contain wrong characters; check every line.
03 · Template direction
Packaging and label concepts for product mockups.
A frosted amber glass bottle on a travertine plinth with olive branches and soft linen, warm Mediterranean sunlight. The cream label reads "NORTH & CO" and below it "DAILY SERUM". English text only.
Not a print-ready label; rebuild final brand text and legal copy in a design tool.
04 · Template direction
Menus, price boards, and signs mixing Chinese and English.
A chalkboard menu hanging in a cozy cafe. Title: "TODAY'S MENU 今日菜单". Three lines below: "Latte 拿铁 ¥28", "Matcha 抹茶 ¥32", "Croissant 可颂 ¥18". Hand-drawn chalk style, warm light.
Check every price and character before using the image.
Use the defaults as a starting point, then choose the model and settings that fit the actual image task.
Describe the subject, layout, lighting, and style. Put each exact word or line of text in quotation marks.
Choose a canvas ratio. MorphPix runs MOSS Image at its 1K resolution tier.
Sign in, check the Credit cost, generate, then proofread every word before you use the image.
Task guidance
Facts below come from the official OpenMOSS GitHub repository and Hugging Face model cards. MorphPix does not publish its own benchmark for this model.
MOSS-Image is a 5B-parameter text-to-image diffusion transformer from the OpenMOSS team, released on October 16, 2026 as MOSS-Image-Base and MOSS-Image-SFT. The weights and bundled inference code are Apache-2.0. OpenMOSS highlights text rendering, layout control, and prompts of up to 4K tokens.
Describe the scene first, then list every piece of text in quotation marks and say where it goes, for example top headline, bottom line, or a label on the bottle. Locally, the official pipeline also accepts a JSON layout with bounding boxes; this page sends plain-text prompts.
FLUX.2 Pro on MorphPix handles both text-to-image and reference-guided edits from an uploaded image. MOSS Image is text-to-image only, but its release focuses on long rendered text and layout. Pick MOSS Image for text-heavy posters and signs; pick FLUX.2 Pro when you need to edit a photo.
Qwen-Image is Alibaba's open-weight model, also known for Chinese and English text rendering, and is available on MorphPix's Qwen Image page. Both target text-rich images; MOSS Image adds JSON layout input when run locally and is a smaller 5B model. Compare both on your own prompt before choosing.
OpenMOSS says ComfyUI supports MOSS Image natively: update ComfyUI to a master or nightly build with MOSS Image support, download the ComfyUI-format weights from huggingface.co/Comfy-Org/MOSS-Image, and use the built-in loaders. Full steps are in the comfyui folder of github.com/OpenMOSS/MOSS-Image. The full checkpoint needs a large NVIDIA GPU; use this page if you do not have one.
OpenMOSS (Fudan University)
long Chinese and English text inside images, posters, signs, packaging concepts, and text-heavy layouts
20 credits
The model weights are free to download under Apache-2.0. Generating on MorphPix uses Credits: each successful 1K image costs the Credits shown before you generate, and failed jobs are not charged. The Free plan includes a monthly Credit allowance.
Yes, that is its main strength. Put the exact text in quotation marks. Long or small text can still contain mistakes, so proofread the result.
No. MOSS Image is a text-to-image model. Use the AI Image Editor on MorphPix to change an existing photo.
MorphPix runs the 1K tier (about one megapixel) in the aspect ratio you choose.
The official code is at github.com/OpenMOSS/MOSS-Image, the weights are on Hugging Face under OpenMOSS-Team/MOSS-Image-Base and OpenMOSS-Team/MOSS-Image-SFT, and OpenMOSS also offers an API on the MOSI platform.
The model weights are Apache-2.0. Before commercial use, also check MorphPix's terms and the rights to any logos, people, or brands in your image.
Open the studio with the prompt, model, style, and aspect ratio already filled in.
AI Poster Generator
Describe the poster purpose, visual subject, headline direction, palette, and style to generate a poster concept from text. Proofread important copy before publishing.
AI Image Generator
Describe the image you want, choose from nine current AI generation models, set a style and aspect ratio, then sign in to review the Credit cost before generating.
AI Logo Generator
Describe your brand name, symbol, typography direction, style, and palette to explore raster logo concepts. Proofread every letter and rebuild the selected direction as vector artwork before final use.
YouTube Thumbnail Generator
Describe the video topic, focal subject, visual hook, headline area, and color direction to create a 16:9 AI thumbnail concept from text. Proofread and replace important lettering before publishing.