Метаверсище и ИИще

Вы будете смеяться, но у нас новый генератор картинок. Точнее foundation model для оного.

Опенсорсный, с кодом, веса бахнули сегодня.

1. Text-to-Image

2. ID customization

3. Multiview generation

Text to multiview

4. Condition-to-Image and vice versa

5. Subject-driven generation

6. Text-guide image editing

7. Zero-shot Task combinations

https://github.com/lehduong/OneDiffusion

Щас его упихают в Комфи, а пока там Омнигеновские требования к памяти:

The demo provides guidance and helps format the prompt properly for each task. By default, it loads the Molmo for captioning source images, which significantly increases memory usage. You generally need a GPU with at least 40 GB of memory to run the demo. Opting to use LLaVA can reduce this requirement to about ≈27 GB, though the resulting captions may be less accurate in some cases.

Всем удачных тестов!

@cgevent

www.tgoop.com/cgevent/10502

6.7K viewsSergey Tsyptsyn ️️, Dec 11 at 14:34

Вы будете смеяться, но у нас новый генератор картинок. Точнее foundation model для оного.

Опенсорсный, с кодом, веса бахнули сегодня.

1. Text-to-Image

2. ID customization

3. Multiview generation

Text to multiview

4. Condition-to-Image and vice versa

5. Subject-driven generation

6. Text-guide image editing

7. Zero-shot Task combinations

https://github.com/lehduong/OneDiffusion

Щас его упихают в Комфи, а пока там Омнигеновские требования к памяти:

The demo provides guidance and helps format the prompt properly for each task. By default, it loads the Molmo for captioning source images, which significantly increases memory usage. You generally need a GPU with at least 40 GB of memory to run the demo. Opting to use LLaVA can reduce this requirement to about ≈27 GB, though the resulting captions may be less accurate in some cases.

Всем удачных тестов!

@cgevent

BY Метаверсище и ИИще