MOSS-Image
MOSS-Image is a 5 billion parameter text-to-image model from the OpenMOSS team, released on 16 October 2026 in two checkpoints, a base model and a supervised fine-tune, MOSS-Image-SFT, which is the one fal hosts. It specialises in rendering Chinese and English text and in structured layout: a prompt can be plain language, tags, or a JSON layout that places text and objects in boxes. On fal, prompt expansion is on by default and turns a plain prompt into exactly such a layout, with a box for every element, before the image is drawn. The pipeline code is under Apache 2.0. There is no edit endpoint.
Anvisha Pai, Co-founder & CEO, Voyager
OpenMOSS · released
›What we asked for
A self-portrait of MOSS-Image, an open image model built for text and layout: a moss-covered stone garden sign in a misty forest, the words 'MOSS IMAGE' carved cleanly into the stone, a smaller line beneath reading 'text and layout', ferns around the base, soft morning light.
fal-ai/moss-image
image_size: {"width":1360,"height":1024}, output_format: png, num_images: 1
returned 1360 by 1024 pixels
At a glance
- Price
- $0.05 a megapixel on fal: $0.05 for a 1024 square, about $0.066 at 1024 by 1360 and $0.083 for our 2016 by 864 banner.
- Sizes
- Any width and height from 512 to 2048 pixels, rounded to multiples of 16. We used 1024 square, 1024 by 1360, 1360 by 1024 and 2016 by 864.
- Layout planning
- With prompt expansion on, the default, the prompt is rewritten as a JSON layout with a box for every text line and object, returned as actual_prompt. You can also write that layout yourself.
- Open weights
- Base and SFT checkpoints from OpenMOSS; the pipeline code is Apache 2.0.
- No editing
- Text-to-image only on fal, so the edit prompt below could not be run.
Anvisha's take
This is the model to watch from the October batch. Its layout planner turned our chart prompt into a JSON plan with a box for every bar, gridline and label, and the chart came out right: six values on a correct 0 to 300 axis, bars drawn to them, April's 170 below March's 180. No other October addition managed that. The poster, logo and five-object composition were also exactly right, three ice cubes and all, and the logo is a genuine merge of chain link and spoked wheel with no tagline. The planner also likes to add things: an invented subtitle and paragraphs on the infographic, with a typo ('eliminisate'), a 'Users (k)' unit on the chart, and an archway framing the market. At five cents a megapixel it is not the cheapest, and the 21:9 banner took a minute. But for text and layout, an open model this literal is new.
MOSS-Image real output: the same 12 prompts every model gets
Every image model on this site gets the same 12 prompts, so you can compare like with like. These are the pictures MOSS-Image returned, one try each, nothing re-rolled or retouched. Generated on . The 11 pictures took 14 to 63 seconds each and cost $0.6979 in total.
Spelling, kerning and hierarchy of three lines of set text.
›What we asked for
A concert poster that reads 'MODA SUMMER SESSIONS' in large bold type, with the subtitle 'Friday 12 June, 8pm, Roundhouse' and a smaller line 'Tickets at the door'. Risograph style, two colours, tangerine and navy, grainy ink texture.
fal-ai/moss-image
image_size: {"width":1024,"height":1360}, output_format: png, num_images: 1
returned 1024 by 1360 pixels
Catalogue lighting, a clean seamless background and a believable contact shadow.
›What we asked for
Studio product photograph of a matte black ceramic pour-over coffee dripper on a seamless white background, soft top light, subtle contact shadow, centred, catalogue style.
fal-ai/moss-image
image_size: {"width":1024,"height":1024}, output_format: png, num_images: 1
returned 1024 by 1024 pixels
A natural face, correct finger count and hands holding an object.
›What we asked for
Close-up portrait of a woman in her 40s with short grey hair, laughing, holding a red ceramic mug with both hands, natural window light, 85mm lens, shallow depth of field.
fal-ai/moss-image
image_size: {"width":1024,"height":1360}, output_format: png, num_images: 1
returned 1024 by 1360 pixels
A flat single-colour mark with a legible wordmark and no gradients or photo texture.
›What we asked for
A flat vector logo mark for a bicycle repair shop called 'Spoke & Chain': a single-colour icon combining a chain link and a wheel spoke, with the wordmark below in a geometric sans-serif, black on white.
fal-ai/moss-image
image_size: {"width":1024,"height":1024}, output_format: png, num_images: 1
returned 1024 by 1024 pixels
Many distinct objects and people in one frame without merging or duplicating them.
›What we asked for
A crowded Saturday farmers market seen from slightly above: at least twelve stalls with striped awnings, shoppers carrying tote bags, a dog on a leash, crates of oranges, a musician with a guitar, bunting overhead, late morning light.
fal-ai/moss-image
image_size: {"width":1360,"height":1024}, output_format: png, num_images: 1
returned 1360 by 1024 pixels
A named historical style applied faithfully rather than a generic painterly look.
›What we asked for
A lighthouse on a rocky cliff during a storm, in the style of a Japanese ukiyo-e woodblock print, with visible wood grain texture, flat colour areas and a bokashi gradient sky.
fal-ai/moss-image
image_size: {"width":1024,"height":1360}, output_format: png, num_images: 1
returned 1024 by 1360 pixels
Six labelled bars at the stated heights, in order, with readable axes.
›What we asked for
A clean bar chart titled 'Monthly active users' showing six months, January through June, with the values 120, 150, 180, 170, 220 and 260, labelled axes, white background, one accent colour, no 3D effects.
fal-ai/moss-image
image_size: {"width":1360,"height":1024}, output_format: png, num_images: 1
returned 1360 by 1024 pixels
A 21:9 composition that leaves the left third clear for headline text.
›What we asked for
Website hero banner: a wide panorama of a sunlit modern co-working space with plants, the left third of the frame kept as clear wall space for text, warm neutral palette.
fal-ai/moss-image
image_size: {"width":2016,"height":864}, output_format: png, num_images: 1
returned 2016 by 864 pixels
Four numbered steps in the given order with their exact wording, plus an icon each.
›What we asked for
An infographic card titled 'How to brew pour-over' with four numbered steps in this order: 1. Rinse the filter, 2. Bloom for 30 seconds, 3. Pour in slow circles, 4. Wait 3 minutes. A simple icon beside each step, muted palette, clean sans-serif type.
fal-ai/moss-image
image_size: {"width":1024,"height":1360}, output_format: png, num_images: 1
returned 1024 by 1360 pixels
Appetising, menu-grade food with correct textures and no melted anatomy.
›What we asked for
Overhead photograph of a margherita pizza on a wooden board, fresh basil leaves, one slice pulled out with the cheese stretching, restaurant menu photography, natural light.
fal-ai/moss-image
image_size: {"width":1024,"height":1024}, output_format: png, num_images: 1
returned 1024 by 1024 pixels
Exact counts and left-right placement of five named objects.
›What we asked for
A glass of water on a wooden table with exactly three ice cubes in it, a yellow pencil resting across the rim of the glass, a blue book to the left of the glass and a red apple to the right, photographed from table height.
fal-ai/moss-image
image_size: {"width":1360,"height":1024}, output_format: png, num_images: 1
returned 1360 by 1024 pixels
Change one object's colour and add a part while leaving background, light and framing untouched.
›What we asked for
Change the coffee dripper to a deep forest green and add a small brass handle on its right side. Keep the background, lighting and composition exactly as they are. (applied to the product on white result above)
fal-ai/moss-image
How the cost under each picture was worked out: fal's MOSS-Image rate of $0.05 per megapixel for 1024x1360, 1.33 MP at 1024x1024 per MP; fal does not state rounding (fal.ai/models/fal-ai/moss-image, read 2026-10-22) and does not return the charge.
Compare MOSS-Image
MOSS-Image settings that matter
Inference steps
Steps barely matter for type: both posters and both infographics spelled every line correctly at 20 and at 50, the default. 20 steps was faster, 10.6 against 18.8 seconds on the poster and 19.3 against 23.4 on the infographic, at the same price. Each run plans a fresh layout, so the 50-step pictures here differ from the 50-step samples above; the infographics still gained invented sub-lines either way.
Text in an image
›What we asked for
A concert poster that reads 'MODA SUMMER SESSIONS' in large bold type, with the subtitle 'Friday 12 June, 8pm, Roundhouse' and a smaller line 'Tickets at the door'. Risograph style, two colours, tangerine and navy, grainy ink texture.
fal-ai/moss-image
image_size: {"width":1024,"height":1360}, num_inference_steps: 20, output_format: png, num_images: 1
returned 1024 by 1360 pixels
Infographic with steps
›What we asked for
An infographic card titled 'How to brew pour-over' with four numbered steps in this order: 1. Rinse the filter, 2. Bloom for 30 seconds, 3. Pour in slow circles, 4. Wait 3 minutes. A simple icon beside each step, muted palette, clean sans-serif type.
fal-ai/moss-image
image_size: {"width":1024,"height":1360}, num_inference_steps: 20, output_format: png, num_images: 1
returned 1024 by 1360 pixels
MOSS-Image pricing
| Where | Unit | Price | Source |
|---|---|---|---|
| fal | Megapixel fal does not say how it rounds; our row costs are the exact output megapixels at 1024 by 1024 per megapixel. | $0.05 | fal model page, read |
Is MOSS-Image free?
The checkpoints are downloadable to run yourself. Hosted on fal it is paid per megapixel.
- •Self-hosted: base and SFT checkpoints and the pipeline from OpenMOSS.
- •fal: $0.05 a megapixel.
MOSS-Image on GitHub, read
Where to use MOSS-Image
- GitHub · Pipeline code, Apache 2.0, and links to the checkpoints.
- fal · fal-ai/moss-image, the SFT model, per megapixel; the endpoint used for the samples below.
- Voyager · Runs inside Voyager, the agent for creative work, with your own fal key.
Voyager
Run MOSS-Image inside Voyager
Voyager is an agent for creative work that runs every model in this reference, MOSS-Image included. Brief it, and it picks the model, makes the thing and hands back something you can edit. Create an account to get started.
What we noticed running MOSS-Image
- •Correct: six values on a 0 to 300 axis with bars drawn to them, April below March. Months are abbreviated and it added a 'Users (k)' unit. Its returned layout plan has a box for every bar, gridline and label. See the chart picture.
- •Book left, apple right, three ice cubes and the pencil across the rim, though everything sits small in the frame. See the hard composition picture.
- •A chain link merged into a spoked wheel, single colour, with a spaced 'SPOKE & CHAIN' wordmark and nothing else. See the logo mark picture.
- •All three lines correct and nothing added, with a two-colour sun, soundwave and road graphic. See the text in an image picture.
- •Four headings exact with icons, plus an invented subtitle and a sentence under every step, one with a typo ('eliminisate'). See the infographic with steps picture.
- •The left third is a bare, sunlit wall with no text, at 2016 by 864. It took 63 seconds, the slowest request in the run. See the wide aspect ratio picture.
- •Dense market with a seated busker, his amp and a golden retriever on a lead, framed through a stone archway nobody asked for. See the busy scene picture.
- •A ribbed cone with a handle on a base, catalogue-clean, though small and off-centre in the frame. See the product on white picture.
- •Most pictures took 14 to 30 seconds at the default 50 steps.
- •The self-portrait carved 'MOSS-IMAGE', with a hyphen it was not given, and 'text and layout' cleanly into the stone.
MOSS-Image limits and API parameters
| Limit | Value |
|---|---|
| Output size | 512 to 2048 pixels per side, in multiples of 16. fal API schema, read |
| Edits | None: fal hosts text-to-image only. fal model page, read |
| Licence | The pipeline code is Apache 2.0; the repository does not state the weights' licence on its front page, so check the model cards before commercial self-hosting. MOSS-Image on GitHub, read |
›Every setting, for developers
fal (fal-ai/moss-image)
fal API schema, read
| Parameter | Type | Values | Default |
|---|---|---|---|
| prompt | string | Up to 16,384 characters: natural language, tags or a JSON layout Required. | |
| enable_prompt_expansion | boolean | true, false Plans a structured layout from the prompt, keeping requested text, at the chosen size. | true |
| image_size | enum or object | square_hd, square, portrait_4_3, portrait_16_9, landscape_4_3, landscape_16_9, or {width, height} Rounded to multiples of 16 between 512 and 2048. | square_hd |
| num_inference_steps | integer | 1 to 100 | 50 |
| guidance_scale | number | 0 to 10 | 3 |
| negative_prompt | string | Any text Empty by default, when guidance skips a transformer layer instead. | |
| num_images | integer | 1 to 4 | 1 |
| output_format | enum | jpeg, png, webp | jpeg |
| seed | integer | Any integer |
What MOSS-Image will not draw
On fal a safety checker is on by default and can only be turned off with account authorisation; images it flags come back black. fal marks the endpoint for commercial use. OpenMOSS publishes no separate content policy that we found.
fal API schema, read
Alternatives to MOSS-Image
Frequently asked questions
What is MOSS-Image?
An open 5 billion parameter text-to-image model from OpenMOSS, released on 16 October 2026, built for accurate Chinese and English text and structured layouts, which it plans as JSON before drawing.
How much does MOSS-Image cost?
$0.05 a megapixel on fal, read on 22 October 2026: five cents for a 1024 square image.
Can MOSS-Image edit images?
Not on fal, which hosts text-to-image only.
What is the JSON layout?
A structured prompt listing each text line and object with a bounding box. With prompt expansion on, MOSS writes one from your plain prompt and returns it; you can also write your own to place elements exactly.
How this page is made
Prices, limits and settings are read from the linked vendor and host pages on the dates shown. Every picture was generated by us through fal with the request shown under it, and the original files are kept unedited. A generation that failed is shown as a failure. Model and vendor names are used to identify the products; no vendor imagery appears on this page.
