Kimi K3Kimi K3 MOSS-Image Workspace for Text, Posters and Visual Ideas
Use MOSS-Image SFT inside the independent Kimi K3 workspace. This page runs the OpenMOSS image model through fal; it does not claim that Kimi K3.1 generates these still images. Write a scene, include English or Chinese lettering, or plan a composition with JSON and inspect official source examples.
Image workspace
Sample Image
3 / 40Official OpenMOSS Base sample. Reuse the prompt with the selected model; results and sampling settings differ.
Introduction
Kimi K3: Meet MOSS-Image, the OpenMOSS Image Model
MOSS-Image brings written descriptions and structured composition into the same generation workflow. Explore an open model family designed for lettering, illustration, photography and graphics with several distinct elements.
a compact generator with room for detail
MOSS-Image uses a five-billion-parameter diffusion backbone with a language encoder that interprets a detailed visual brief. Describe the subject, setting, materials and intended reading order together.
choose the checkpoint you actually use
The model family supports English and Chinese lettering as part of the picture, making it useful for fictional notices, classroom scenes and poster concepts. This workspace runs the SFT version through fal.
MOSS-Image: published examples and model identity
MOSS-Image Base is the original foundation; SFT adds supervised fine-tuning on curated data while keeping the shared generation interface. Official showcase pictures below identify their Base origin, so you can distinguish published examples from results made here.
Official OpenMOSS Base examples; this workspace uses SFT through fal. The overview video presents still examples, not generated motion. Model source · Prompt guide · Video introduction
How to use
How to Create with MOSS-Image
- STEP 1
Describe your MOSS-Image idea
Choose text-to-image and write the subject, scene and visual style. Put any visible wording in quotation marks. For a poster, explain the title hierarchy and where the main illustration belongs. You can also load a published example and adapt its prompt.
- STEP 2
Set the canvas and composition
Select a ratio and longest-edge preset, or enter supported custom dimensions. If you use a JSON layout, keep its coordinates consistent with the canvas. MOSS-Image prompt expansion can help develop a short brief; turn it off for a layout you already prepared.
- STEP 3
Review the MOSS-Image quote
Choose one to four pictures and your output format. Advanced controls include steps, guidance, a random or fixed seed, and a negative prompt. Check the displayed credits, sign in and submit when the inputs describe the result you want.
- STEP 4
Inspect and continue creating
Open the result and read every important word. Compare the subject, object positions and crop with your brief. Download a useful picture, revise the prompt for another attempt, or take your MOSS-Image asset into the included photo, design and document tools.
Features
Kimi K3: Put English and Chinese Words into MOSS-Image Pictures
Use MOSS-Image to connect wording with the scene around it. A headline on a poster, handwriting on a note and print on a textbook each need different visual instructions.
write the visible text explicitly
Place the exact phrase inside quotation marks, then describe its material, color, size and location. MOSS-Image can work with a short English headline beside Chinese supporting copy.
give letters a physical setting
Specify line breaks when the division matters, particularly for signs and narrow labels. Keep product names and dates easy to inspect. Handwriting on a chalkboard should respond to its perspective and surface, while a clean poster can use regular typography and larger margins.
MOSS-Image: lettering in the finished design
Tell MOSS-Image which treatment belongs in your scene. Compare the finished wording with your requested text, including punctuation, repeated characters and missing lines.
Kimi K3: Compose with MOSS-Image Layout Prompts
A layout prompt describes the overall scene and individual elements together. MOSS-Image accepts structured JSON so a title, illustration and smaller text block can each receive a planned position.
arrange the important elements first
Start with a clear hierarchy: one main title, one visual subject and a supporting line. A common MOSS-Image layout uses a summary plus an items array, with descriptions and bounding boxes for text and objects.
keep composition editable in your brief
In the documented centered coordinate format, boxes use vertical and horizontal limits measured from the canvas center. Change an element description, its wording or its box, then generate a new picture from that specification. MOSS-Image layout control guides synthesis; it does not lock pixels from a previous image.
MOSS-Image: a brief that preserves your intent
Turn off prompt expansion when you already have a complete JSON layout. For a short natural-language idea, expansion can instead plan a structured composition.
Kimi K3: Explore MOSS-Image across Illustration and Photography
Typography is only one part of the model. MOSS-Image also responds to descriptions of paper sculpture, drawn characters, everyday photography and cinematic environments, giving your brief several useful visual directions.
MOSS-Image: describe material rather than decoration
For a paper-cut scene, explain the folded layers, visible edges, depth and arrangement on a desk. MOSS-Image can interpret these material relationships without a long string of quality adjectives.
MOSS-Image: use light to establish the scene
For a painted illustration, describe brush texture, palette and the scale of the subject. The official rainy-street example combines warm shop light with cooler reflections, a transparent umbrella and wet pavement. Use MOSS-Image to develop your own setting with similarly concrete relationships.
MOSS-Image: a scene with believable relationships
State where the light comes from and which subject it reveals. Describe the viewpoint, weather and distance between objects.
Kimi K3: Match MOSS-Image Settings to the Finished Asset
Choose the shape and dimensions for the destination before submitting. MOSS-Image here supports one to four outputs, selectable formats and advanced sampling controls, with the quote shown beside the generation action.
Choose dimensions with an actual purpose
The 1K, 1.5K and 2K presets describe the longest edge; the selected ratio determines the other edge. MOSS-Image also accepts custom dimensions from 512 to 2048 pixels on each side, aligned to multiples of sixteen. A square graphic and a wide banner therefore have different pixel areas even with the same preset label.
Understand what changes the credit quote
Layout coordinates should use those actual dimensions rather than an assumed square canvas. The published fal rate is five cents per million output pixels. MOSS-Image generation converts that amount at one cent per credit, rounds up for each picture, then multiplies by the output count.
MOSS-Image: sampling controls and output choices
Steps, guidance, seed and output format do not add a separate charge in this pricing rule. The interface and server share the calculation. Review the displayed total before submitting; subscription and credit-pack purchase prices remain visible in the pricing dialog.
Kimi K3: Take a MOSS-Image Concept into Your Creative Workflow
Use the generated picture as one asset within a larger project. MOSS-Image can supply a first illustration, a typographic composition or a visual direction that you continue in the included creative tools.
Move from an example to your own brief
The example gallery includes published images with their full prompts and source-model labels. Load a prompt into MOSS-Image, change its subject or wording, and choose supported settings for your own request. Loading an example only prepares the workspace.
Keep creation and later editing clear
It does not submit a generation. For detailed layout examples, preserve the corresponding dimensions and disable expansion before editing positions, so the composition remains understandable as you develop it. MOSS-Image SFT in this workspace takes text and produces images. It does not accept a reference photograph for editing, and it does not generate video.
MOSS-Image: the next step for your asset
Other model selections and the included creative editors support separate tasks. After generation, open a result in the appropriate tool to add approved lettering, assemble a document or prepare a video reference. Each new AI request shows its own supported inputs and credits.
Example prompts
Bilingual coffee poster
A clean cream-and-emerald coffee poster with a ceramic cup in the lower half. At the top, large dark green text reads "MORNING RITUAL". Below it, smaller Chinese text reads "把清晨留给自己". Generous margins, clear typography and soft window light.
Paper sculpture story
An open pop-up book on a wooden desk, revealing a layered paper ocean with turquoise waves, coral-red boats and a small lighthouse. Visible folded paper edges, realistic shadows between layers, front-facing composition, soft daylight, no lettering.
User reviews
Kimi K3 video creator feedback.
Kimi K3 creator reviews
Illustrative workflow: plan an original game title sequence with simple shapes, readable words and timed visual beats. Inspect the animation before adding it to a larger game presentation.
Illustrative workflow: turn a short explanation into a diagram that reveals one idea at a time. Compare Sketch and Motion treatments, then keep the version that communicates the subject clearly.
Illustrative workflow: place an approved product picture beside short benefit text. Assemble a simple graphic product story and review the composition at the intended screen size.
Illustrative workflow: create a portrait opening with one headline, a small set of colors and a quiet final hold. Export the selected scene and add separate sound during editing when needed.
Illustrative workflow: revise one visual decision in a saved animation, such as the closing pause or label placement. Compare saved versions before exporting the selected result.
Illustrative workflow: build a silent graphic sequence to explain the structure of a presentation. These examples describe possible uses; they are not verified customer testimonials or promised outcomes.
Kimi K3 Pricing and Credit Plans
Cancel anytime
Lite
50% off0.025 / credit
Billed $ yearly179· Save $179.8
Video Models
Image Models
- Includes Kimi K3.1 and all premium models
- Up to 1 batch generation task
- Standard generation speed
- Standard generation success rate
- Standard customer support
- Commercial Use License
Standard
50% off0.017 / credit
Billed $ yearly299· Save $299.8
Video Models
Image Models
- Includes Kimi K3.1 and all premium models
- Up to 4 batch generation tasks
- Priority processing speed
- High generation success rate
- Priority customer support
- Commercial Use License
Pro
50% off0.014 / credit
Billed $ yearly599· Save $599.8
Video Models
Image Models
- Includes Kimi K3.1 and all premium models
- Up to 10 batch generation tasks
- Fastest generation speed
- High generation success rate
- Dedicated account manager
- Commercial Use License
Max
50% off0.012 / credit
Billed $ yearly1,199· Save $1,199.8
Video Models
Image Models
- Includes Kimi K3.1 and all premium models
- Up to 10 batch generation tasks
- Fastest generation speed
- High generation success rate
- Dedicated account manager
- Commercial Use License
Pay safely and securely with
FAQ
Frequently asked questions
What is MOSS-Image?
MOSS-Image is an open text-to-image model family developed by OpenMOSS. It supports natural-language descriptions, tag prompts and structured layouts, with particular attention to English and Chinese text inside images. This site runs SFT through fal.
Does MOSS-Image support Chinese lettering?
Yes. Include the Chinese wording exactly and describe its location, size and line breaks. The official examples include textbook pages and notices. Generated characters can still differ from your request, so inspect any text that matters before publishing.
Can MOSS-Image edit an uploaded photograph?
The integrated checkpoint is text-to-image only. Choose a separate editing model or open a creative editor for existing pictures. Changing a layout prompt generates a new composition; it does not preserve pixels from a previous result.
How much does MOSS-Image generation cost?
The quote uses the actual pixel area at the published five-cent-per-megapixel rate, converted at one cent per credit and rounded up per image. A 1024 by 1024 image costs six credits; four such outputs cost twenty-four credits.
Are the gallery pictures made with MOSS-Image SFT?
The imported official showcase records identify Base as their source model. Their prompts are available as starting points for the selected SFT endpoint, whose output can differ. Branding artwork and explanatory videos are identified separately from model-generated examples.
Does MOSS-Image create videos or guarantee exact text?
The current integration produces still images. The presentation video demonstrates published still examples, rather than model-generated motion. Neither lettering nor layout is guaranteed to match every instruction; check the result in its intended destination.
CTA
Start Your Next MOSS-Image Composition
Bring a phrase, a scene or a layout into the workspace. Choose MOSS-Image SFT, set the canvas and turn a clear visual idea into an image you can inspect and develop.
Create with MOSS-Image