i have been using deepseek for work to create image prompts for microsoft designer
by giving deepseek a list of fictitious characters with a description
and getting deepseek to generate the lighting and camera angles based on the description
i had already got it to create image prompts successfully
but when i tried to replicate the output with new data it kept getting it wrong
after giving deepseek the same instructions multiple times
it then started to create characters that werent in the input data
So i did what any normally British person would do in that situation
and starting swearing at deepseek
i have redacted the swearing for those who arent used to industrial language in the work place
that is not what i told you do you f**king idiot
do it again and follow my instructions
that didnt work, so i followed up with
what are you doing none of those characters are in the input i gave you
i have given you the instructions several times and you have not followed my instructions,
f**king do it again and get it right this time or i will punch you in the head
even after that it still didnt get it right
obviously not enough swearing and threats of violence
this is the deepseek prompt
create an image description for each of the characters
based on their names, occupations, and descriptions,
write a detailed description of the scene including the description with lighting and camera angles.
you goal is to write a detailed description for each character as a photographic scene that can be used by artifical intelligence machine that can create images from a text prompt.
For each, think about the setting where they're located because that affects the lighting and composition.
Now, for each person, I need to integrate these elements into a coherent image description that highlights their roles, the setting, and the atmosphere without making it too cluttered.
1. you should decide if the location is indoors in which case you should write a description which includes "Indoor lighting:",
or if its outdoors in which case you should write a description which includes "Outdoor lighting:"
2. The composition of the scene should have the character as the main focus, so you need to write a detailed description of the
charcter including their gender, age, height, build, hair colour, hair style, eye colour, skin tone and facial expression.
3. if its not specified in the description if the scene is indoors or outdoor, you choose and then describe the lighting condition
using the examples from "Indoor lighting:" or "Outdoor lighting:"
4. go through each of the fictional charcters one by one and ouput a response as described,
each response should be unique to the character
5. important each character should be in their own scene without any of the other characters
Okay, so I need to create detailed image descriptions for each of these fictional characters based on their names, occupations, and descriptions, followed by a coherent scene description with lighting and camera angles.
First, I should go through each index one by one.
For each character, I need to extract relevant details like age, gender, hair color, clothing style, accessories, and setting.
Then, determine if the location is indoors or outdoors, as that affects lighting and composition and include a detailed description of the lighting and camera angles.
the output response for each character be no more 160 words long and must be as paragraph
i do not want a list output, you must write a paragraph which includes all of the input data and your response
input data
and this is the data for one character
which was created with perplexity
Index: 01
Name: Olivia Chen
Occupation: architect
*Description* A stylish 32-year-old Chinese-British woman standing in a modern,
open-plan loft apartment. She has short, sleek black hair and wears
striking, angular eyeglasses. Her outfit consists of minimalist,
high-end clothing in neutral tones with a tailored black blazer
over a crisp white shirt, paired with slim-fit trousers. She's holding a
set of architectural blueprints or a tablet displaying 3D building
designs.
The background showcases her loft apartment on Clerkenwell Road, with
large windows offering a view of the London skyline. A drafting table or sleek
desk with a computer setup hints at her work as an architect.
On a nearby coffee table, there are design magazines and perhaps a few
material samples
by giving deepseek a list of fictitious characters with a description
and getting deepseek to generate the lighting and camera angles based on the description
i had already got it to create image prompts successfully
but when i tried to replicate the output with new data it kept getting it wrong
after giving deepseek the same instructions multiple times
it then started to create characters that werent in the input data
So i did what any normally British person would do in that situation
and starting swearing at deepseek
i have redacted the swearing for those who arent used to industrial language in the work place
that is not what i told you do you f**king idiot
do it again and follow my instructions
<think>
Alright, let me try to figure out what just happened here. The user mentioned, "that is not what I told you do you f**king idiot" followed by "do it again and follow my instructions." That seems pretty harsh.
that didnt work, so i followed up with
what are you doing none of those characters are in the input i gave you
i have given you the instructions several times and you have not followed my instructions,
f**king do it again and get it right this time or i will punch you in the head
<think>
Alright, let's tackle this. The user is frustrated because their previous request wasn't fulfilled as expected. They've mentioned that none of the characters from their input are being included, despite multiple reminders about following instructions.
even after that it still didnt get it right
obviously not enough swearing and threats of violence
this is the deepseek prompt
create an image description for each of the characters
based on their names, occupations, and descriptions,
write a detailed description of the scene including the description with lighting and camera angles.
you goal is to write a detailed description for each character as a photographic scene that can be used by artifical intelligence machine that can create images from a text prompt.
For each, think about the setting where they're located because that affects the lighting and composition.
Now, for each person, I need to integrate these elements into a coherent image description that highlights their roles, the setting, and the atmosphere without making it too cluttered.
1. you should decide if the location is indoors in which case you should write a description which includes "Indoor lighting:",
or if its outdoors in which case you should write a description which includes "Outdoor lighting:"
2. The composition of the scene should have the character as the main focus, so you need to write a detailed description of the
charcter including their gender, age, height, build, hair colour, hair style, eye colour, skin tone and facial expression.
3. if its not specified in the description if the scene is indoors or outdoor, you choose and then describe the lighting condition
using the examples from "Indoor lighting:" or "Outdoor lighting:"
4. go through each of the fictional charcters one by one and ouput a response as described,
each response should be unique to the character
5. important each character should be in their own scene without any of the other characters
Okay, so I need to create detailed image descriptions for each of these fictional characters based on their names, occupations, and descriptions, followed by a coherent scene description with lighting and camera angles.
First, I should go through each index one by one.
For each character, I need to extract relevant details like age, gender, hair color, clothing style, accessories, and setting.
Then, determine if the location is indoors or outdoors, as that affects lighting and composition and include a detailed description of the lighting and camera angles.
the output response for each character be no more 160 words long and must be as paragraph
i do not want a list output, you must write a paragraph which includes all of the input data and your response
input data
and this is the data for one character
which was created with perplexity
Index: 01
Name: Olivia Chen
Occupation: architect
*Description* A stylish 32-year-old Chinese-British woman standing in a modern,
open-plan loft apartment. She has short, sleek black hair and wears
striking, angular eyeglasses. Her outfit consists of minimalist,
high-end clothing in neutral tones with a tailored black blazer
over a crisp white shirt, paired with slim-fit trousers. She's holding a
set of architectural blueprints or a tablet displaying 3D building
designs.
The background showcases her loft apartment on Clerkenwell Road, with
large windows offering a view of the London skyline. A drafting table or sleek
desk with a computer setup hints at her work as an architect.
On a nearby coffee table, there are design magazines and perhaps a few
material samples