Image Generation

Sprelay exposes POST /v1/images/generations. The combination gpt-image-2, 1024x1024, low, and Base64 output has been tested with a temporary low-quota key. Image editing is not documented because it has not been verified.

Before you begin

1. Confirm gpt-image-2 in the Model Catalog.
2. Choose the current dedicated image group when creating the key.
3. Give that key a separate, small quota.
4. Generate one low-quality image for the first test.

curl

# Save the large JSON response to a file instead of filling the terminal.
curl https://sprelaytoken.com/v1/images/generations \
  -H "Authorization: Bearer $SPRELAY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-image-2",
    "prompt": "A simple flat blue circle centered on a white background.",
    "n": 1,
    "size": "1024x1024",
    "quality": "low"
  }' \
  --output response.json

prompt describes the image, n: 1 requests one result, and size plus quality use the tested low-cost combination. JSON cannot contain comments.

Decode the first image on a system with jq and base64:

# Extract Base64 from JSON, decode it, and write a PNG file.
jq -r '.data[0].b64_json' response.json | base64 --decode > sprelay-image.png

Python

# base64 decodes the image; requests sends HTTP; os reads the key.
import base64
import os
import requests

# Image generation can take longer than text, so use a 180-second timeout.
response = requests.post(
    "https://sprelaytoken.com/v1/images/generations",
    headers={
        "Authorization": f"Bearer {os.environ['SPRELAY_API_KEY']}",
        "Content-Type": "application/json",
    },
    json={
        "model": "gpt-image-2",
        "prompt": "A simple flat blue circle centered on a white background.",
        "n": 1,
        "size": "1024x1024",
        "quality": "low",
    },
    timeout=180,
)

# Stop on 4xx/5xx before trying to decode an error response.
response.raise_for_status()
payload = response.json()
image_bytes = base64.b64decode(payload["data"][0]["b64_json"])

# wb writes the decoded bytes as a binary PNG file.
with open("sprelay-image.png", "wb") as image_file:
    image_file.write(image_bytes)

PowerShell

# Read the key from the current PowerShell environment.
$headers = @{
  Authorization = "Bearer $env:SPRELAY_API_KEY"
  "Content-Type" = "application/json"
}

# Request one tested low-quality image to control the first-call cost.
$body = @{
  model = "gpt-image-2"
  prompt = "A simple flat blue circle centered on a white background."
  n = 1
  size = "1024x1024"
  quality = "low"
} | ConvertTo-Json

# PowerShell parses the JSON response automatically.
$response = Invoke-RestMethod `
  -Uri "https://sprelaytoken.com/v1/images/generations" `
  -Method Post `
  -Headers $headers `
  -Body $body `
  -TimeoutSec 180

# Decode the first Base64 image and save it in the current directory.
$bytes = [Convert]::FromBase64String($response.data[0].b64_json)
[IO.File]::WriteAllBytes("sprelay-image.png", $bytes)

Base64 image data is large. Decode or upload it promptly; do not write the complete value to application logs. Before retrying a timeout, check usage logs because the upstream image may already have been generated and charged.