How to Generate and Edit Images Using OpenAI gpt-image-1 API

OpenAI's latest multimodal model, gpt-image-1, revolutionizes image generation within ChatGPT and via its API. This article explores its features, usage, and applications.

Table of Contents

• Understanding gpt-image-1
  • Key Capabilities of gpt-image-1
  • Accessibility
• gpt-image-1 Pricing Structure
• Accessing gpt-image-1
• Practical Application of gpt-image-1
  • Image Generation with gpt-image-1
  • Image Editing with gpt-image-1
• Best Practices for Optimal Results
• Diverse Applications of gpt-image-1
• Limitations of gpt-image-1
• Model Comparison
• Conclusion

Understanding gpt-image-1

gpt-image-1 is OpenAI's advanced multimodal model, generating high-quality images informed by real-world knowledge. While gpt-image-1 is recommended, the Image API also supports DALL·E 2 and DALL·E 3.

The Image API provides three key functions:

• Image Creation: Generate images from text prompts.
• Image Modification: Edit existing images using new prompts.
• Image Variations: Generate variations of an existing image (DALL·E 2 only).

Also Read: A Comparison of Imagen 3 and DALL·E 3

Key Capabilities of gpt-image-1

gpt-image-1 boasts:

• High-fidelity output: Detailed and accurate visuals.
• Versatile styles: Supports diverse aesthetics, from photorealistic to abstract.
• Precise editing: Allows targeted image modifications.
• Contextual understanding: Interprets complex prompts accurately.
• Reliable text rendering: Consistently renders text within images.

Accessibility

OpenAI's API allows image generation and editing using GPT Image or DALL·E models. Currently, image generation is exclusively via the Image API, with Responses API support under development.

For more details on gpt-image-1, click here.

gpt-image-1 Pricing Structure

Understanding the pricing is crucial for efficient and cost-effective usage.

gpt-image-1 is priced per token:

• Text prompts: $5 per 1M tokens
• Uploaded images: $10 per 1M tokens
• Generated images: $40 per 1M tokens

This translates to approximately:

• ~$0.02 for a low-quality image
• ~$0.07 for a medium-quality image
• ~$0.19 for a high-quality image

Detailed pricing information (by image quality and resolution) is available on the official pricing page here.

Note: Image generation involves creating specialized tokens, impacting both processing time and cost. Larger images and higher quality settings increase both.

Accessing gpt-image-1

To obtain your gpt-image-1 API key:

1. Log in to the OpenAI platform.
2. Navigate to Project > API Keys.
3. Verify your account (KYC verification may be required).

Visit https://www.php.cn/link/c7f409723f7655153ec02c23b87b31a9 for account verification details.

gpt-image-1: Practical Application

Let's explore image generation and editing using the gpt-image-1 API.

Image Generation Using gpt-image-1

We'll use the image generation endpoint with text prompts. The API defaults to a single image, but the 'n' parameter allows generating multiple images per request.

Before running the main code, set up the environment:

<code>!pip install openai
import os
os.environ['OPENAI_API_KEY'] = "<your-openai-api-key>"</your-openai-api-key></code>

Input Code (Example):

from openai import OpenAI
import base64
client = OpenAI()

prompt = """A serene park scene with humans and friendly robots enjoying the day."""

result = client.images.generate(model="gpt-image-1", prompt=prompt)

image_base64 = result.data[0].b64_json
image_bytes = base64.b64decode(image_base64)

with open("park_scene.png", "wb") as f:
    f.write(image_bytes)

Output:

Image Editing Using gpt-image-1

gpt-image-1 allows for various image edits:

• Editing existing images.
• Using other images as references.
• Inpainting (editing specific image areas using a mask).

Editing with a Mask (Example):

Transparent mask areas are replaced based on the prompt; colored areas remain unchanged.

Input Code (Example - adding Elon Musk):

from openai import OpenAI
client = OpenAI()

result = client.images.edit(
    model="gpt-image-1",
    image=open("/content/image.png", "rb"),
    mask=open("/content/mask.png", "rb"),
    prompt="Elon Musk added to the image"
)

# ... (rest of the code to save the image remains the same)

Output:

(Note: Replace /content/image.png and /content/mask.png with your actual file paths.) Remember image and mask must be same format, dimensions, and under 25MB. A black and white mask needs an alpha channel added.

Best Practices for Optimal Results

1. Customize image settings (size, quality, format, compression).
2. Use square images (1024x1024) and standard quality for speed.
3. The API returns base64-encoded data; default format is PNG, but JPEG or WebP are also options.
4. For JPEG/WebP, control compression with output_compression (0-100%).

Diverse Applications of gpt-image-1

gpt-image-1 has broad applications across various fields: gaming, creative tools, education, enterprise software, advertising, healthcare, architecture, and entertainment.

Limitations of gpt-image-1

• Latency: Complex prompts can take up to 2 minutes.
• Text Rendering: While improved, text clarity and alignment can be inconsistent.
• Consistency: Maintaining uniformity across multiple images can be challenging.
• Composition Control: Precise element placement may not always be perfect.

Model Comparison

Model	Endpoints	Features
DALL·E 2	Generations, Edits, Variations	Lower cost, concurrent requests, inpainting
DALL·E 3	Generations	Higher resolution, better quality than DALL·E 2
gpt-image-1	Generations, Edits (Responses API coming soon)	Excellent instruction-following, detailed edits, real-world awareness

Conclusion

gpt-image-1 offers powerful image generation and editing capabilities from text prompts. While processing time can vary, its quality and control make it a valuable tool. It facilitates faster content creation and prototyping. While enhancing human creativity, it’s important to consider ethical implications and the value of original human work.

The above is the detailed content of How to Generate and Edit Images Using OpenAI gpt-image-1 API. For more information, please follow other related articles on the PHP Chinese website!