DALL-E API Integration for Image Generation

DALL-E API Integration for Image Generation

DALL-E 3 from OpenAI is the most controllable model for business use cases: precisely follows text instructions, understands complex composite prompts, requires no GPU infrastructure. API is available through OpenAI SDK.

Core Features

from openai import AsyncOpenAI
import base64
import httpx

client = AsyncOpenAI()

async def generate(prompt: str, **kwargs) -> bytes:
    response = await client.images.generate(
        model="dall-e-3",
        prompt=prompt,
        size=kwargs.get("size", "1024x1024"),   # 1024×1024, 1792×1024, 1024×1792
        quality=kwargs.get("quality", "standard"),  # standard ($0.04), hd ($0.08)
        style=kwargs.get("style", "vivid"),          # vivid / natural (realistic)
        n=1,  # DALL-E 3 supports only n=1
        response_format="b64_json"
    )
    return base64.b64decode(response.data[0].b64_json)

async def edit_image(image_bytes: bytes, mask_bytes: bytes, prompt: str) -> bytes:
    """Edit existing image (inpainting)"""
    response = await client.images.edit(
        model="dall-e-2",  # Edit API only for DALL-E 2
        image=image_bytes,
        mask=mask_bytes,
        prompt=prompt,
        size="1024x1024",
        n=1,
        response_format="b64_json"
    )
    return base64.b64decode(response.data[0].b64_json)

async def create_variation(image_bytes: bytes) -> bytes:
    """Create variations of existing image"""
    response = await client.images.create_variation(
        model="dall-e-2",
        image=image_bytes,
        n=1,
        size="1024x1024",
        response_format="b64_json"
    )
    return base64.b64decode(response.data[0].b64_json)

Prompt Engineering for DALL-E 3

DALL-E 3 rewrites the prompt internally (automatic enhancement). To track it:

response = await client.images.generate(
    model="dall-e-3",
    prompt=prompt,
    # System instruction through "I NEED to..." for exact following
)
# Actually used prompt (after internal processing):
print(response.data[0].revised_prompt)

Effective templates:

# Commercial photography
"{product}, professional commercial photography, white background, studio lighting, 8k, product shot"

# Article illustration
"flat illustration of {concept}, modern minimal style, pastel colors, no text"

# Banner
"{subject}, banner composition, horizontal format, space for text on left side, professional"

FastAPI Integration

from fastapi import FastAPI
from pydantic import BaseModel

app = FastAPI()

class ImageRequest(BaseModel):
    prompt: str
    size: str = "1024x1024"
    quality: str = "standard"

@app.post("/generate")
async def generate_image(request: ImageRequest):
    image_bytes = await generate(request.prompt, size=request.size, quality=request.quality)
    filename = f"{uuid.uuid4()}.png"
    url = await upload_to_s3(filename, image_bytes)
    return {"url": url, "filename": filename}

Cost: standard 1024×1024 — $0.040, HD — $0.080, 1792×1024 HD — $0.120. For 10,000 images per month — $400–800. For high loads, self-hosted FLUX or SDXL is more economical. Integration timeline — 1–3 days.