Gpt-image-1 b_64json response incomplete or corrupted

API Bugs
1

I am testing the new gpt-image-1 API to do simple editing, this is my code:

// Always add product image
  const productImageResponse = await fetch(payload.productImageUrl);
  const productImageBlob = await productImageResponse.blob();
  const productImageFile = new File([productImageBlob], "product-image.png", { type: productImageBlob.type });
  
  logger.log("Product image file:", productImageFile);

  // Generate the image using OpenAI with gpt-image-1 model
  const result = await openai.images.edit({
    model: "gpt-image-1",
    prompt: payload.prompt,
    n: 1,
    size: "1024x1024",
    image: productImageFile,
  });

it runs well, but the response’s b64_json seems to be either incomplete or something as I cant even decode it to view the image, this is the response:

{
“created”: 1745519700,
“data”: [
{
“b64_json”: “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”
}
],
“usage”: {
“input_tokens”: 296,
“total_tokens”: 4456,
“output_tokens”: 4160,
“input_tokens_details”: {
“text_tokens”: 102,
“image_tokens”: 194
}
}
}

2

Welcome to the community.

Is it giving an error when you try to decode? All requests come back the same?

1 Like
3

Yep, I’ve tried both on the image edit and generate image

4

Okay. It’s working here…

How are you decoding it?

1 Like
5

const result = await openai.images.edit({
model: “gpt-image-1”,
prompt: payload.prompt,
n: 1,
size: “1024x1024”,
image: productImageFile,

  // Get the image data
  const imageBase64 = result.data[0].b64_json;

   const { data, error } = await supabase
.storage
.from("product-images")
.upload(`${productId}/${fileName}`, Buffer.from(imageBase64, "base64"), {
    contentType: "image/png",
    upsert: false
  });

ahh, may I know how are u decoding it?

6

I’ve also tried throwing the b64_json data into a decoder online Base64 to Image | Base64 Decode | Base64 Converter | Base64,

A webpage interface for converting Base64 text to an image with fields for input and file information, but the image preview cannot be displayed in the browser. (Captioned by AI)
A webpage interface for converting Base64 text to an image with fields for input and file information, but the image preview cannot be displayed in the browser. (Captioned by AI)1724Ă—764 145 KB

7

The b64_json also seems a little bit short to be an image, no?

Sorry in advance for newbie question

8

Aight, it’s solved, my bad, it was due to other technology that trigger the function

1 Like