How can I correctly upload a local image to GPT‑4 Vision API?

Below is an explanation and a revised code example that uses base64 encoding. Instead of reading the image with matplotlib (which produces an array), you need to open the file in binary mode, encode it, and then pass it as a data URI string in the "url" field.

Steps to Load a Local Image for GPT‑4 Vision

1. Read and Encode the Image:
   Open the image file in binary mode and encode it using base64.

2. Create a Data URI:
   Format the encoded string as a data URI with the correct MIME type (for example, image/png).

3. Send the Request:
   Pass the resulting data URI as the "url" in your message payload.

Example Code

PYTHONCopied
import base64
from openai import OpenAI

client = OpenAI()

# Function to encode the image as base64
def encode_image(image_path):
    with open(image_path, "rb") as image_file:
        return base64.b64encode(image_file.read()).decode('utf-8')

# Path to your local image
image_path = "img.png"
base64_image = encode_image(image_path)

# Construct the data URI string (update MIME type if needed)
data_uri = f"data:image/png;base64,{base64_image}"

response = client.chat.completions.create(
    model="gpt-4-vision-preview",
    messages=[
        {
            "role": "user",
            "content": [
                {"type": "text", "text": "What’s in this image?"},
                {
                    "type": "image_url",
                    "image_url": {
                        "url": data_uri,
                    },
                },
            ],
        }
    ],
)

print(response.choices[0].message.content)

Explanation

• Base64 Encoding:
  The encode_image function reads your local file and converts it into a base64-encoded string.

• Data URI Construction:
  The encoded image is prefixed with data:image/png;base64, to form a data URI. Make sure to use the correct MIME type (e.g., image/jpeg if your image is a JPEG).

• Sending the Request:
  The created data URI is then used in the API request, enabling GPT‑4 vision capabilities to analyze your local image.

This approach directly aligns with the best practices for passing local image data to the OpenAI API for image processing.