Comfui images vs Offical demo about Lumina-Image 2.0
- Dominant language
- Python
- Stars
- 133k
- Forks
- 15.7k
- Avg merge
- 1d 6h
- Merged PRs (30d)
- 155
Description
Thanks to ComfyUI for supporting Lumina-Image 2.0. We have noticed that the images generated by ComfyUI differ significantly from those produced by our official demo, as shown below:
| Prompt | ComfyUI Images | Official Demo |
|----------------------------------|------------------------------------------|-----------------------------------------|
| 一位穿着深蓝色盔甲的神秘剑士特写,他的面容隐没在金属面罩后,眼中透露着冷酷的光芒,背后是闪烁的城市灯火,剑身上闪现着红色能量,面罩上有复杂的纹路,环境暗淡、充满肃杀气氛 |  |  |
| A charming half-body portrait of a female with deep eyes and long curly hair, soft light outlines her delicate facial features. Her skin is delicate, with a smile on the corner of her mouth, and her eyes are gentle but firm. Her right hand gently stroked her collarbone, her fingertips were slender and elegant, and her hand lines were smooth and natural. A beam of sunlight shone on her face, sharp light and shadow, the background is soft warm tones to blur the light and shadow,|  |  |
|This is a high-definition indoor photo with a style that leans towards daily life In the picture, there is a young woman wearing a gray tank top and black sports shorts, sitting on the floor with her legs slightly bent and her body facing to the right Her eyes were fixed on the camera in her hand, as if carefully observing or adjusting the camera settings Her expression was calm, her hair was flowing down, dark brown in color, with long hair flowing down and no bangs She wears a white wristband on her left hand and holds a camera in her right hand On her left side, there is a wooden table with a glass of water and a pair of headphones placed on it In the background, some blurry furniture and items can be seen, such as white plush toys and wooden cabinets |  |  |
| A charismatic speaker is captured mid-speech. He has short, tousled brown hair that's slightly messy on top. He has a round circle face, clean shaven, adorned with rounded rectangular-framed glasses with dark rims, is animated as he gestures with his left hand. He is holding a black microphone in his right hand, speaking passionately. The man is wearing a light grey sweater over a white t-shirt. He's also wearing a simple black lanyard hanging around his neck. |  |  |
|充满魔法光辉的城市,夕阳下,拥有粉色长发的魔法少女穿着华丽的战斗服,手持发光权杖,背景是未来都市的摩天大楼,光影交错,充满梦幻色彩|  |  |
| Layered paper of wave, aesthetic, epic detail, dreamy blue color, starry night |  |  |
The potential reasons for the differences are:
- **CFG Renorm is vital** for improving generation quality.
- **System prompt is missing**, which plays an important role in enhancing the visual quality and instrution following ability.
- **CFG truncation is not implemented**, acceleration without compromising quality.
These factors contribute significantly to the variations observed between ComfyUI and our official demo outputs.
Contributor guide
Assessment
This issue has not been assessed yet.