VibeusLoadingЗагрузка
Vibeus

Loading is taking longer than usual.Загрузка занимает больше времени, чем обычно.

The full page couldn't load.Не удалось полностью загрузить страницу.

You can read the available content below or try loading the page again.Вы можете читать доступный контент ниже или повторить загрузку страницы.

MOSS-Image is open source: the team has released the code and... - Vibeus

Maya Chen ·

MOSS-Image is open source: the team has released the code and weights for its 5B DiT image model.

Its core developer says the project took six months and focused on three areas: rendering text inside images, controlling composition through layouts, and leaving a flexible base for community fine-tuning. The official repo describes two variants, Base and SFT, plus support for prompts up to 4K tokens, anime and photorealistic styles, and fine-tuning for custom tasks. It also includes a layout-generation example and a visual comparison of the two variants.

In an AMA reply, the developer recommends Base for LoRA training and SFT for everyday use; Base tends to be more photorealistic, while SFT may have stronger aesthetic quality. The team is answering questions about the release and says it will share details openly.

Источник

Publication imagePublication imagePublication image