Difference between Describe with ai and tag
First, create the prompt words.
Finally, generate images for each prompt word.
- ①Original image…Original image for testing
- ②I'm GETags image…Generated using app tags
- ③Describe image…Images generated using tag words that describe the original image
- ④Remixed image…Click the remix button on the original image to generate
①Original Image
I used this illustration to look up the tag.
Prompt words for the original image
face details,semi realistic, Masterpiece, Master work, perfect , 4k, 1woman, wet hair, mature body, medium size ?, straight half body long hair, middle parted hair style, large forehead, black colored hair, golden eyes, arrogant expression face, wrapped white towel, collarbone, ? feet, standing pose, showing wet armpits, empty background, white background, high res, ultrasharp, 8K, masterpiece, looking at the viewer (masterpiece), best quality, expressive eyes, perfect face
This is the original image
Source:https://www.seaart.ai/explore/detail/ct31335e878c7384euhg?source=share_link&u_code=DA8ELQRE
②AI Apps I'm GETags.
Generate prompt words, Generate tags with images
How the app works
APP-generated prompts
1girl, solo, long hair, ?s, looking at viewer, blush, simple background, shirt, large ?s, black hair, white background, closed mouth, cleavage, ? shoulders, medium ?s, brown eyes, collarbone, white shirt, upper body, yellow eyes, armpits, mole, arms up, parted bangs, eyelashes, wet, lips, ? arms, makeup, towel, wet clothes, arms behind head, forehead, water drop, red lips, adjusting hair, wet hair, ? towel, dripping, hands in hair, parted hair, white towel,
app generated image and text(This is the original image)
AI Apps I'm GETags. : https://www.seaart.ai/workFlowAppDetail/cvshqjte878c73cjlkm0
③Describe with ai
Describe the original image to generate a word.
When read an image with Describe with AI, it will generate prompt words.
Now you have two prompt words.
- anime girl in a white dress posing in the water, ? anime girl, beautiful anime girl, attractive anime girl, guweiz, beautiful anime woman, beautiful alluring anime woman, beautiful anime portrait, pretty anime girl, clean detailed anime art, artwork in the style of guweiz, beautiful alluring anime , detailed portrait of anime girl, detailed digital anime art
- wet long hair, draped in a white cloth, water droplets on skin, relaxed pose, natural smooth skin, outdoor setting with bright lighting, strong highlights and shadows, fresh and clean atmosphere
The result was the following image. The seed was not fixed. I remixed it as it was and generated the image.
Now have three prompt words.
For comparison, I wrote the original prompt first.
0 Original
face details,semi realistic, Masterpiece, Master work, perfect , 4k, 1woman, wet hair, mature body, medium size ?, straight half body long hair, middle parted hair style, large forehead, black colored hair, golden eyes, arrogant expression face, wrapped white towel, collarbone, ? feet, standing pose, showing wet armpits, empty background, white background, high res, ultrasharp, 8K, masterpiece, looking at the viewer (masterpiece), best quality, expressive eyes, perfect face
1 app
1girl, solo, long hair, ?s, looking at viewer, blush, simple background, shirt, large ?s, black hair, white background, closed mouth, cleavage, ? shoulders, medium ?s, brown eyes, collarbone, white shirt, upper body, yellow eyes, armpits, mole, arms up, parted bangs, eyelashes, wet, lips, ? arms, makeup, towel, wet clothes, arms behind head, forehead, water drop, red lips, adjusting hair, wet hair, ? towel, dripping, hands in hair, parted hair, white towel,
2 Describe
anime girl in a white dress posing in the water, ? anime girl, beautiful anime girl, attractive anime girl, guweiz, beautiful anime woman, beautiful alluring anime woman, beautiful anime portrait, pretty anime girl, clean detailed anime art, artwork in the style of guweiz, beautiful alluring anime , detailed portrait of anime girl, detailed digital anime art
3Describe
wet long hair, draped in a white cloth, water droplets on skin, relaxed pose, natural smooth skin, outdoor setting with bright lighting, strong highlights and shadows, fresh and clean atmosphere
Now, use each prompt word in the Original condition.
1 app Prompt
1girl, solo, long hair, ?s, looking at viewer, blush, simple background, shirt, large ?s, black hair, white background, closed mouth, cleavage, ? shoulders, medium ?s, brown eyes, collarbone, white shirt, upper body, yellow eyes, armpits, mole, arms up, parted bangs, eyelashes, wet, lips, ? arms, makeup, towel, wet clothes, arms behind head, forehead, water drop, red lips, adjusting hair, wet hair, ? towel, dripping, hands in hair, parted hair, white towel,
2 Describe
anime girl in a white dress posing in the water, ? anime girl, beautiful anime girl, attractive anime girl, guweiz, beautiful anime woman, beautiful alluring anime woman, beautiful anime portrait, pretty anime girl, clean detailed anime art, artwork in the style of guweiz, beautiful alluring anime , detailed portrait of anime girl, detailed digital anime art
3 Describe
wet long hair, draped in a white cloth, water droplets on skin, relaxed pose, natural smooth skin, outdoor setting with bright lighting, strong highlights and shadows, fresh and clean atmosphere
About these simple results.
dis ai's second "Wet long black hair~" is the closest to the original.
The first one is closer to a sentence than a word.
app has moved away from Original. It feels like it was added to Original because of the increased details.
Next
Generate Original with remix without changing the settings
Remixed image
Can perform a remix using a button like this.
The result was an image like this. The seed was not fixed. I just remixed it and generated the image.
Prompt word(Everything remains original)
face details,semi realistic, Masterpiece, Master work, perfect , 4k, 1woman, wet hair, mature body, medium size ?, straight half body long hair, middle parted hair style, large forehead, black colored hair, golden eyes, arrogant expression face, wrapped white towel, collarbone, ? feet, standing pose, showing wet armpits, empty background, white background, high res, ultrasharp, 8K, masterpiece, looking at the viewer (masterpiece), best quality, expressive eyes, perfect face
AI Apps I'm GETags.
I'm GETags. Generate prompt words, Generate tags with images
app: https://www.seaart.ai/workFlowAppDetail/cvshqjte878c73cjlkm0
A side note. As an application,
when I created a LoRAModel and generated an image,
sometimes the face would not appear.
I created four images and the face would not appear in any of them, or in only one.
Perhaps the face was not included because the rectangular image was cropped to a square in the LoRA learning settings.
I don't know the cause,
but when I generated it the face did not appear.
So I tried to get the tags of the image used in the training data with im getags.
In other words, after creating LoRA Model, I was unsure which words to use to display it correctly, so I got all the tags written on the image.
I used that as the prompt word when generating the LoRA. As a result, the face appeared successfully.
I will explain this separately.
