MOSS Image SFTBeta

MOSS Image SFT: AI Image Generation Online

Create images with the fine-tuned MOSS Image SFT model. Explore visual directions for posters, editorial concepts and creative work from a text prompt.

6 credits / imageYour credits work across all models.

Your output

A little input. A new possibility.

Your finished audio, image or text will appear here.

01 / Inside the model

A considered prompt. A new visual direction.

Create images with the SFT variant of the MOSS-Image family. The official project offers Base and SFT checkpoints for text-to-image generation, including bilingual text rendering and varied visual styles. Use the SFT workspace to test a detailed creative direction with a chosen canvas shape. Keep the prompt specific: explain what should appear, how it should be arranged and the mood the image should convey.

Prompt-driven scenes

Specify objects, relationships and the intended visual atmosphere.

Bilingual visual ideas

Explore image concepts involving English or Chinese text.

Compare model variants

Try the same brief with Base to understand the difference in your results.

02 / Examples & listening room

See it. Hear it. Get a feel for it.

Explore the official model

Official examples are credited to OpenMOSS. Studio samples were generated by us with the named API model. These are saved previews, not a result of your current settings. Research capabilities and available studio controls may differ.

SFT checkpoint · Chinese text rendering
Official OpenMOSS example

A stylised portrait
Official OpenMOSS example

03 / Use cases

Where it fits.

Campaign concepts

Develop a hero-image direction or a product storytelling scene.

Illustration drafts

Explore a defined art direction before selecting the final composition.

04 / Questions

A few things worth knowing.

How much does Image SFT cost?

6 credits per generated image, regardless of which preset canvas shape you select.

Does SFT guarantee better output than Base?

No. Results depend on your prompt and task. Compare actual images instead of assuming one variant always wins.

Can I edit an uploaded image?

This workspace currently supports text-to-image generation. Image-to-image editing is not available.

Made for your next idea

Choose a model, add your input and review the credit cost. Successful results stay in your account history; AI output may be inaccurate.

Text is rounded up to a whole credit. Audio is billed per started minute. Live voice uses one credit per 6 seconds.

Failed tasks are refunded automatically.

Result links can expire. Download files you want to keep.

Model documentation