Z-image turbo-New trained - v2

Z-image turbo-New trained

LORA
Original - TenStar Fund


Обновлено:

Recommended settings:heunpp2+linear_quadratic,cfg:1-2,step:8-20,0.6-1 weight

Most effective style change keywords: semi-realistic style, traditional Chinese style, digital painting, black and white photography, 2D anime style, Japanese film style, influencer look, professional photography.

Most effective style change keywords: 半写实风格,中式古风,数字油画,黑白摄影,2D动漫风格,日系胶片风格,网红脸,专业摄影.

This LoRA delivers superior performance across styles including 2D, animation, illustration, film photography, and movie stills. It is compatible with a wide range of aesthetics, with the final effect fully determined by your prompts. Trained using a new method, it comes with both advantages and drawbacks: the benefits include higher efficiency and lower training computing cost, while the downside is potential interference between multi-modal training datasets. However, this may actually be beneficial for creating novel hybrid styles. Moving forward, tests will be conducted using this training approach on single-style datasets.

You can also try out more styles and gameplay on your own.

Have fun!

Детали версии

Trained by Tensor
Z-Image-Turbo
30
heunpp2+linear_quadratic,cfg:1-2,step:8-20,0.6-1 weight

Права доступа к проекту

Перепечатка строго запрещена

    Использовать разрешения

  • Использовать на TENSOR Online

  • Как базовая модель для онлайн-обучения на TENSOR

  • Использовать без указания моего авторства

  • Поделитесь объединениями этой модели

  • Используйте разные разрешения при слияниях

    Коммерческое использование

  • Продажа сгенерированного контента

  • Использовать в сервисах генерации

  • Продать эту модель или объединить

Похожие записи