image_gen.py

import base64

from openai import OpenAI
from dotenv import load_dotenv
import os

from helpers import encode_image, generate_rep_image

load_dotenv()

client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))

def get_image(product, wear, product_category, user_inputs, dropdown_list):

    analyses = client.responses.create(
        model="gpt-4o",
        input=[
            {
                "role": "user",
                "content":[
                    {
                        "type": "input_text",
                        "text": f"Analyse the design and wearing way for the given {product_category}"
                    },
                    {
                        "type": "input_image",
                        "image_url": f"data:image/jpeg;base64,{encode_image(product)}"
                    },
                    {
                        "type": "input_image",
                        "image_url": f"data:image/jpeg;base64,{encode_image(wear)}"
                    }
                ]
            }
        ]
    )

    imgs = []
    for u_input, d_input in zip(user_inputs, dropdown_list):
        prompt_gen = client.responses.create(
        previous_response_id=analyses.id,
        model="gpt-4o",
        input=[
            {
                "role": "user",
                "content": [
                    {
                        "type": "input_text",
                        "text": f"Generate me a prompt for the {product_category}, based on the user request for the given {product_category} which should include following things:"
                                f"User Request: {u_input},"
                                "Things to include:"
                                f"""[1] CORE ─ Subject + Action + Product
                                <Describe who/what, what they’re doing, and the product’s key look>
                                [2] CAMERA ─ Angle / POV / Lens / Crop
                                [3] MOOD ─ Lighting / Colour Grade / Backdrop
                                [4] QUALITY ─ Resolution / Realism / Photographic Style
                                [5] EXTRAS ─ Skin Texture / Makeup / Wardrobe / Atmosphere / Aspect Ratio
                                 Make sure you only return the prompt and should be in single paragraph.
                                 Don't write anything about the design of {product_category} in the prompt.
                                 Where ever you need to mention about the {product_category}, only write '{product_category} in reference image' or similar to this."""
                    }
                ],
            },
        ]
        )
        prompt = prompt_gen.output[0].content[0].text
        print(prompt)

        if d_input == "On Model":
            rep_image = generate_rep_image(wear, prompt)
            image_bytes = base64.b64decode(rep_image)
            imgs.append(image_bytes)
            image_gen = client.responses.create(
                previous_response_id=prompt_gen.id,
                model="gpt-4o",
                input=[
                    {
                        "role": "user",
                        "content": [
                            {
                                "type": "input_text",
                                "text": f"""Create Image for the following request:
                                {prompt}"""
                            },
                            # {
                            #     "type": "input_image",
                            #     "image_url": f"data:image/jpeg;base64,{encode_image(product)}"
                            # },
                            # {
                            #     "type": "input_image",
                            #     "image_url": f"data:image/jpeg;base64,{encode_image(wear)}"
                            # }
                        ]
                    }
                ],
                tools=[
                    {
                        "type": "image_generation",
                        "quality": "high",
                        "size": "1024x1024"
                    }
                ]
            )

            rep_image = generate_rep_image(wear, prompt)
            image_bytes = base64.b64decode(rep_image)
            imgs.append(image_bytes)

        else:
            image_gen = client.responses.create(
                previous_response_id=prompt_gen.id,
                model="gpt-4o",
                input=[
                    {
                        "role": "user",
                        "content": [
                            {
                                "type": "input_text",
                                "text": f"""Create Image for the following request:
                                {prompt}
                                Ensuring the {product_category} remains in natural size and detail, not magnified. The focus is realistic: not zoomed-in or stylized"""
                            },
                            # {
                            #     "type": "input_image",
                            #     "image_url": f"data:image/jpeg;base64,{encode_image(product)}"
                            # }
                        ]
                    }
                ],
                tools=[
                    {
                        "type": "image_generation",
                        "quality": "high",
                        "size": "1024x1024"
                    }
                ]
            )


        image_generation_calls = [
            output
            for output in image_gen.output
            if output.type == "image_generation_call"
        ]

        image_data = [output.result for output in image_generation_calls]

        if image_data:
            image_base64 = image_data[0]

            image_bytes = base64.b64decode(image_base64)
            imgs.append(image_bytes)
            print("Image Generated")
        else:
            print("No image generated")


    return imgs