Good prompt for realistic images (photo, render...)

#1
by Shinku - opened

Update: The first one was for cities only, sorry, this is the same prompt with the city part removed: 這張照片採用了寬廣的景深,並利用自然光營造出充滿活力的氛圍。

Prompt for cities:
Just add ",採用寬廣的景深,捕捉了城市天際線的宏偉,利用自然光營造出充滿活力的街景氛圍。" to the end of your prompt when you're creating something realistic... I was translating prompts from English to Traditional Chinese when Command R just added it by itself... and it worked.

The word "clock" with this prompt added:
image - 2024-07-06T174102.839.webp

Others images with prompts I don't remember using this:
image - 2024-07-06T172948.565.webp

image - 2024-07-06T172643.735.webp

image - 2024-07-06T173136.840.webp

Thanks for wonderful suggestions!

Thanks for wonderful suggestions!

Thanks to you for the space!

These negative are working well too: 模糊,過度壓縮,雜訊,過度處理,曝光不足或過度曝光,失真,像素化,色彩失真,圖像瑕疵,對焦問題

prompt: a laptop in a dark room, leds, 採用寬廣的景深,捕捉了城市天際線的宏偉,利用自然光營造出充滿活力的街景氛圍。
image - 2024-07-06T180029.452.webp

I updated the post, it looks like the prompt was forcing a city.

The translation of the first prompt would be: "The photograph employs a wide depth of field, capturing the grandeur of the urban skyline and utilizing natural light to create a vibrant ambiance of the city streets."

I removed the city part.

Yeah, I suppose Chinese prompt will work better, since the text encoder ChatGLM is a Chinese LLM.

What hardware needs for this?

What hardware needs for this?

22gb ram, 19gb vram

Will you have an api?

Will you have an api?

I just created space for this model, I am not the owner. You can ask this question on their model page.

What hardware needs for this?

22gb ram, 19gb vram

For those who have interest, actually you can run it with ~8GB vram, although you may have to offload a bit to ram.

Sign up or log in to comment