MOSS Image prompts: text, posters and layouts
17 prompts for MOSS Image, the open-source model built for words in pictures. The four with a picture we ran ourselves, first try. "Use this prompt" puts any of them in the tool below, 1 credit an image.
Size
about 1 megapixel · 1 credit
Made by GetStepSnap with MOSS Image, first try. Both lines spelled right, the dot between them too. The band name went to the top on its own.
Your own image will show up here.
Poster prompts
Name the kind of print, two colors at most, and put each line of text in quotes with where it goes.
Made by GetStepSnap with MOSS Image
Gig poster
A screen-printed gig poster in orange and navy for a band called "THE LATE TRAINS", a night train crossing a bridge under a full moon, and the line "FRIDAY 9 PM · THE BLUE ROOM" along the bottom
Both lines spelled right, the dot between them too. The band name went to the top on its own.
Travel poster
A vintage travel poster for "AMALFI" in big cream letters across the top, pastel houses stacked on a cliff above a deep blue sea, lemon trees in the front, flat colors and soft paper grain, with "SUMMER 2027" small along the bottom
Swap the town and the three things it's known for. Keep the year short.
Film poster
A moody poster for a made-up thriller called "THE LAST FERRY": a lone ferry in thick fog at night with one lit window, the title in tall thin letters across the middle, and "IN CINEMAS THIS WINTER" small at the bottom
A made-up title on purpose: we don't copy real film posters.
Market poster
A clean modern poster for the "Saturday Farmers Market": a big illustrated basket of vegetables in the middle, the name across the top, and "8 AM – 1 PM · Town Square" underneath, in green and cream
Three pieces of text, each with its own place: that's the most we'd ask for in one prompt.
Sign prompts
Say what the sign is made of, where it hangs and the light, then the words in quotes.
Made by GetStepSnap with MOSS Image
Bakery sign
A hand-painted wooden sign above a small bakery door that reads "Fresh Bread Daily", soft morning light, a bicycle leaning against the brick wall
Right first time, and it added loaves to the window on its own. No stray words anywhere else in the picture.
Diner neon
A red neon sign reading "OPEN 24 HOURS" in the window of a roadside diner at night, rain on the glass, reflections on the wet parking lot
Neon is joined-up glass, so short words in capitals come out cleanest.
Trail sign
A carved wooden trail sign on a post in a pine forest, with two arrow boards: "Lake 2 mi" pointing left and "Summit 5 mi" pointing right, morning fog between the trees
Two pieces of text with a direction each: a small test of where things go.
Wedding seating sign
A wedding sign on a wooden easel that reads "Find your seat" in white script on a sage green board, eucalyptus around the top corners, a garden party softly blurred behind
For just the sign and your words, the AI sign generator does it without a prompt.
Notes, menus and labels
Break the lines yourself, the way you want them on the page. MOSS Image's makers say that's how a line break is set.
Sticky note
A yellow sticky note on a silver laptop lid, handwritten in blue pen: "Call mom Buy milk Finish the slides!"
Three short lines is about right for a note this size.
Café menu
A café chalkboard menu titled "TODAY" in white chalk, with three lines underneath: "Iced latte 4.50" "Banana bread 3.00" "Soup of the day 6.00"
Prices are text too: check every number on the result.
Birthday cake
A round chocolate birthday cake on a kitchen table with "Happy Birthday Maya" piped across the top in white icing, three lit candles, warm evening light
Icing is a hard surface for letters. Keep the name short.
Notebook page
An open notebook on a wooden desk, one page handwritten in blue ink: "Dear future me, drink more water and call your friends." A cup of tea at the corner, soft window light
Handwriting with a comma and a full stop: punctuation is where models slip.
Sticker
A die-cut vinyl sticker of a cheerful cartoon cactus in a pink pot, with the words "STAY SHARP" in bold rounded letters on a banner below it, white border, plain grey background
A plain background makes it easy to cut out afterwards.
Paintings with a title
A long, careful prompt for the picture, then one title in quotes and where it sits, and "No other text."
Made by GetStepSnap with MOSS Image
Silk mountains
A wide landscape painting where the mountains are made of folded blue-green silk and blue brocade, with a few gold threads running along the ridges. A river winds between them, with one small wooden boat and a boatman to show the scale. Soft, restrained morning light; the far hills fade into mist, and the silk weave is visible up close. In the quiet sky at the upper right, the title "SILK MOUNTAINS" in elegant serif letters. No other text.
The title landed exactly where we asked, upper right, and nothing else got written.
Made by GetStepSnap with MOSS Image
Ink and gold
A wide landscape where the main mountain on the left is pressed into black ink paper and black lacquer, with gold leaf left only on a few ridges and sharp folded-paper edges. A river valley opens on the right, with one small wooden boat on the water. Dark, quiet and dramatic. In the empty space at the upper right, the title "INK AND GOLD" in thin gold serif letters. No other text.
Thin gold letters on a dark sky and still spelled right.
Chinese title
A traditional ink wash painting of misty pine mountains and a small pavilion by a still lake, with the Chinese title "山水" in black brush calligraphy and a small red seal at the upper left. No other text.
The prompt in English, the Chinese words in quotes: what MOSS Image's makers recommend.
A JSON layout prompt
MOSS Image can take a layout instead of a sentence: a short description of the whole picture, then each piece of text and each object with a box saying where it goes. You rarely need to write one. Here, every plain prompt is laid out like this before the picture is drawn. This is the layout it wrote for the gig poster above, with the descriptions cut short so it fits the prompt box.
{
"summary": {
"caption": "A two-color screen-printed gig poster in orange and navy: a night train crossing a steel bridge under a full moon.",
"meta": [
"gig poster",
"screen print",
"orange and navy"
],
"canvas": "single"
},
"items": [
{
"label": "text",
"caption": "Bold screen-print letters across the top",
"bbox": [
-512,
-344,
-360,
344
],
"text": "THE LATE TRAINS"
},
{
"label": "object",
"caption": "A big full moon in cream and orange halftone",
"bbox": [
-336,
-152,
-32,
152
]
},
{
"label": "object",
"caption": "A night train with glowing windows on a steel truss bridge",
"bbox": [
-112,
-384,
192,
384
]
},
{
"label": "object",
"caption": "Graphic river waves under the bridge",
"bbox": [
144,
-432,
384,
432
]
},
{
"label": "text",
"caption": "Clean small letters along the bottom",
"bbox": [
432,
-320,
496,
320
],
"text": "FRIDAY 9 PM · THE BLUE ROOM"
}
],
"bbox_coord_space": "abs_centralized",
"bbox_format": "yxyx"
}- bbox is top, left, bottom, right, in pixels from the middle of the picture. This poster is 864 by 1152, so top to bottom runs from -576 to 576. Negative is up and to the left.
- text is the exact words for a box with the label "text". The caption says how they should look.
- meta is a few tags for the overall style.
What we learned from the four we ran
- Every word in quotes came out right, on the first try, including a middle dot and thin gold letters on a dark sky.
- "No other text" worked. Neither painting got a stray word, a signature or a fake logo.
- Where you say, it goes. "Upper right" put both painting titles in the upper right.
- It fills in on its own. Nobody asked for loaves in the bakery window; they're there.
Honestly, four pictures isn't a test of a model, and we'll add results as we run more. For your own words without writing a prompt at all, the AI lettering generator does brush script, neon, retro signs and chalk, and the AI sign generator does shop, welcome, yard and printable signs.
MOSS Image prompts FAQ
What makes a good MOSS Image prompt?
Put the exact words in quotes and say where each piece goes: across the top, on the door, upper right. Break the lines yourself when there's more than one. Write the rest in plain English, and end with "No other text." when you want nothing else written.
Do I need to write JSON layouts?
No. A plain prompt is enough for most pictures: here, MOSS Image turns it into a layout like the one on this page before drawing. Write your own layout when you want each piece in an exact spot.
How do the boxes in a layout work?
Each box is four numbers: top, left, bottom, right, in pixels from the middle of the picture. In a tall 864 by 1152 picture, top to bottom runs from -576 to 576 and left to right from -432 to 432. That's how MOSS Image wrote the layout for our gig poster.
Can MOSS Image write Chinese text?
Yes. Its makers recommend writing the prompt in English and putting the Chinese words in quotes, like the ink painting prompt above.
Did you run all of these prompts?
The four with a picture, yes: on our MOSS Image page, once each, and those are the first results. The others we wrote for this page and haven't run yet, and the layout is a short version of one MOSS Image wrote for our poster.
Do these prompts work in ComfyUI or the Hugging Face demo?
Yes, they're plain text. Results can look different there, because the number of steps, the guidance and whether the prompt gets laid out first all change the picture. Ours ran through fal at its default settings.
Can I try them here?
Yes. "Use this prompt" puts it in the tool at the top with its shape set. Sign in with Google when you generate; an image takes 1 credit. The prompts are yours to copy into any tool.
