GPT-Image-2 API Quickstart: Generate Images with Python in 5 Minutes

TL;DR

• GPT-Image-2 supports up to 4K resolution, transparent PNG output, and image inpainting.
• Connect via CodeGateway: base_url="https://api.codegateway.dev/v1" with your standard OpenAI SDK. Five lines to early image.
• strong for: logos with embedded text, transparent product cutouts, and editing existing images.

GPT-Image-2 Specs

Source: OpenAI API documentation (May 2026). Check OpenAI's pricing page for current rates.

CodeGateway multiplies official prices by the applicable tier rate (1.2x–1.5x). See CodeGateway pricing tiers.

5-Minute Python Quickstart

pip install openai

Basic generation

from openai import OpenAI

client = OpenAI(
    api_key="your-codegateway-api-key",
    base_url="https://api.codegateway.dev/v1"
)

response = client.images.generate(
    model="gpt-image-2",
    prompt="An orange cat sitting on a stack of books, reading under warm lamplight, watercolor illustration style",
    n=1,
    size="1024x1024"
)

image_url = response.data[0].url
print(f"Image URL: {image_url}")  # Valid for 60 minutes

Save to disk (base64)

import base64

response = client.images.generate(
    model="gpt-image-2",
    prompt="Minimalist app UI screenshot, blue theme, clean white background",
    n=1,
    size="1024x1024",
    quality="high",
    output_format="png"
)

img_bytes = base64.b64decode(response.data[0].b64_json)
with open("output.png", "wb") as f:
    f.write(img_bytes)
print("Saved to output.png")

Transparent background (for UI assets, stickers)

response = client.images.generate(
    model="gpt-image-2",
    prompt="Clean app icon representing a code gateway, deep blue, logo style",
    n=1,
    size="1024x1024",
    quality="high",
    output_format="png",
    background="transparent"  # requires PNG or WebP
)

Parameter Reference

size

quality

Cost by quality tier (1024×1024, via CodeGateway)

Generating 100 medium-quality images per day: ~$190/month at new-user rate. Multiplier drops as cumulative spend grows — see tier pricing.

Image Editing (Inpainting)

GPT-Image-2 supports text-guided inpainting — modify part of an existing image while keeping the rest intact.

import base64

with open("original.png", "rb") as f:
    original_b64 = base64.b64encode(f.read()).decode()

with open("mask.png", "rb") as f:
    mask_b64 = base64.b64encode(f.read()).decode()

response = client.images.edit(
    model="gpt-image-2",
    image=original_b64,
    mask=mask_b64,           # White = areas to replace; Black = keep as-is
    prompt="Replace the background with a sunset beach scene",
    n=1,
    size="1024x1024"
)

Mask format: PNG, same dimensions as the original. White pixels mark areas to regenerate; black pixels preserve the original.

Batch Generation (up to 10 images)

response = client.images.generate(
    model="gpt-image-2",
    prompt="Data visualization dashboard, neon blue glow effect, dark background, tech aesthetic",
    n=4,           # 4 variations in one API call
    size="1024x1024",
    quality="medium"
)

for i, img in enumerate(response.data):
    print(f"Variant {i+1}: {img.url}")

GPT-Image-2 vs Imagen 4: Which to Use?

When to pick which:

• Logos, posters, or any image with readable text → GPT-Image-2
• Ultra-realistic people/scenes → Imagen 4 Ultra
• Edit an existing image (inpainting) → GPT-Image-2
• Stylized illustration with no text → comparable; choose by cost or personal preference

Full comparison: Image Generation API Comparison: Imagen, Gemini, and GPT-Image.

FAQ

Q: How long are generated image URLs valid?

A: A: 60 minutes. For persistent storage, request response_format="b64_json" and store the bytes yourself.

Q: Does GPT-Image-2 support non-English prompts?

A: A: Yes, including Chinese and most major languages. English prompts tend to give more precise control over stylistic nuances.

Q: How fast is generation?

A: A: Low quality: 5–10s. Medium: 10–20s. High: 20–40s. Server load and network conditions vary.

Q: Can it generate faces?

A: A: Yes, with OpenAI's standard content policy applied. Generating likenesses of specific real people is rejected.

Q: Can I reuse an OpenAI API key with CodeGateway?

A: A: No — CodeGateway API keys work only with base_url="https://api.codegateway.dev/v1". They're separate credentials.

Related

• Image API Comparison: Imagen, Gemini, and GPT-Image
• Imagen 4 API Quickstart
• CodeGateway Pricing Tiers
• CodeGateway Billing Guide
• OpenAI Images API Reference