Repolode
제출하기
OpenMOSS

OpenMOSS/MOSS-Image

Open-source 5B-parameter text-to-image diffusion model focused on accurate text rendering, JSON layout control, and fine-tuning.

대체할 수 있는 제품:MidjourneyDALL·E

GitHub에서 보기웹사이트
스타
867
포크
42
열린 이슈
7
마지막 커밋
4시간 전

어떤 프로젝트인가요?

MOSS-Image is a 5B-parameter diffusion transformer for text-to-image generation from OpenMOSS. It is released as a base model (MOSS-Image-Base) and a fine-tuned variant (MOSS-Image-SFT) trained on curated data. The project emphasizes generating clear, accurate text inside images and accepts prompts of up to 4K tokens.

Beyond plain prompts, the model accepts JSON layout inputs that specify the placement of text and objects through bounding boxes in pixel coordinates relative to the canvas center. Users can run it from Python through Diffusers with a custom pipeline, or launch a Gradio web interface. A ComfyUI plugin and a prompt-enhancer agent skill are also included.

The repository provides memory modes for GPU-resident, component-offloaded, and sequential inference, along with LoRA fine-tuning guidance in the user guide. Model weights are hosted on Hugging Face and ModelScope, and the code is licensed under Apache-2.0.

핵심 기능

  • 01Generates images with clear, accurate text rendering.
  • 02Controls text and object placement with JSON layout bounding boxes.
  • 03Accepts prompts of up to 4K tokens.
  • 04Runs through Diffusers Python API or a Gradio web interface.
  • 05Provides a ComfyUI plugin and multiple GPU memory modes.
  • 06Supports LoRA fine-tuning for custom tasks.

활용 사례

  • Creating posters or infographics where text must be legible and well placed.
  • Generating images from detailed, long prompts with multiple described elements.
  • Fine-tuning a base diffusion model for a custom style or domain.
  • Building a layout-driven image pipeline with programmatic bounding boxes.

토픽

#text-to-image#diffusion-models#image-generation#text-rendering#layout-control#fine-tuning#pytorch#controllable-generation#foundation-models

자주 묻는 질문

MOSS-Image은(는) 무엇인가요?
Open-source 5B-parameter text-to-image diffusion model focused on accurate text rendering, JSON layout control, and fine-tuning. MOSS-Image is a 5B-parameter diffusion transformer for text-to-image generation from OpenMOSS. It is released as a base model (MOSS-Image-Base) and a fine-tuned variant (MOSS-Image-SFT) trained on curated data. The project emphasizes generating clear, accurate text inside images and accepts prompts of up to 4K tokens.
MOSS-Image은(는) 무료로 사용할 수 있나요? 라이선스는 무엇인가요?
MOSS-Image은(는) Apache-2.0 라이선스의 오픈소스입니다. 상업용 제품에 사용하기 전에 라이선스 조건을 확인하세요.
MOSS-Image은(는) 아직 유지보수되고 있나요?
활발히 유지보수 중. MOSS-Image의 마지막 커밋은 2026년 10월 19일입니다.
MOSS-Image은(는) 어디에 사용할 수 있나요?
Creating posters or infographics where text must be legible and well placed. Generating images from detailed, long prompts with multiple described elements. Fine-tuning a base diffusion model for a custom style or domain. Building a layout-driven image pipeline with programmatic bounding boxes.
MOSS-Image은(는) 어디서 받을 수 있나요?
소스 코드와 설치 방법은 GitHub에 있습니다: https://github.com/OpenMOSS/MOSS-Image

요약은 프로젝트 README를 바탕으로 AI가 작성했습니다. GitHub 통계는 매일 갱신됩니다.

관련 프로젝트