GPT Image 2 API with one flat price per image

1. Create a key

   Sign up, top up from $5 and create a key in API keys. Image models are served in the gpt-value channel group, so use a key on Auto or bound to that group.

2. Point the OpenAI SDK at Lovely Router

   Set the base URL to https://lovelyrouter.com/v1. Requests pass through as OpenAI defines them, so the official SDKs work unchanged.

3. Generate an image

   POST /v1/images/generations holds the connection until the image is ready and returns it as base64 in b64_json.

   cURLPythonNode.js

   Shell

   curl https://lovelyrouter.com/v1/images/generations \
     -H "Authorization: Bearer $LOVELY_API_KEY" \
     -H "content-type: application/json" \
     -d '{
       "model": "gpt-image-2",
       "prompt": "A watercolor painting of a red fox in a snowy birch forest at dawn",
       "size": "1024x1024",
       "quality": "medium"
     }' | jq -r '.data[0].b64_json' | base64 --decode > fox.png

An image usually takes 40 seconds to 2 minutes, plus any time it waits in the queue. The gateway answers within nine minutes, so give your HTTP client a timeout of at least ten. response_format: "url" is ignored; images always come back as base64.

To edit, or to give the model reference images (a product, a logo, a person, a style), send up to 16 images, and optionally a mask, to POST /v1/images/edits with any of the models: as multipart/form-data files, or as URLs in JSON. An edit costs the same as a generation.

curl https://lovelyrouter.com/v1/images/edits \
  -H "Authorization: Bearer $LOVELY_API_KEY" \
  -F model=gpt-image-2 \
  -F image=@fox.png \
  -F prompt="Make it night with a full moon" | jq -r '.data[0].b64_json' | base64 --decode > fox-night.png

Every image request joins a queue and starts as soon as capacity is free, so a burst waits its turn instead of failing. Each account can have up to 20 images queued or generating at once. If you'd rather not hold a connection open for minutes, use image jobs:

# Submit: returns 202 with the job id and its queue_position
curl https://lovelyrouter.com/v1/images/jobs \
  -H "Authorization: Bearer $LOVELY_API_KEY" \
  -H "content-type: application/json" \
  -d '{"model": "gpt-image-2", "prompt": "A red fox in a snowy birch forest", "size": "1536x1024"}'

# Wait up to 60 seconds for it, then download the file
curl "https://lovelyrouter.com/v1/images/jobs/img_8c1f0e2a9b7d4f3e6a5c2b1d?wait=60" \
  -H "Authorization: Bearer $LOVELY_API_KEY"
curl -o fox.png https://lovelyrouter.com/v1/images/jobs/img_8c1f0e2a9b7d4f3e6a5c2b1d/content \
  -H "Authorization: Bearer $LOVELY_API_KEY"

A job takes the same fields as a generation; input images make it an edit. A finished job's image stays available for 7 days. You can list your recent jobs and cancel one that is still queued. If a synchronous request isn't done after nine minutes, it answers 504 with the job id in the x-lovely-image-job header, and the image keeps generating so you can collect it as a job.

To try prompts and edits without code, switch the playground to Images. It runs on image jobs.

Agents can make images too. The MCP server has a generate_image tool that uses GPT Image 2.5, or GPT Image 2 when 2.5 is unavailable, at the same price per image, and takes reference images by URL. It answers with a download link that works for 24 hours without a key, plus a curl command that saves the file; get_image waits for an image that wasn't finished yet. Add the server to Claude Code:

claude mcp add --transport http lovely-tools https://lovelyrouter.com/mcp \
  --header "Authorization: Bearer $LOVELY_API_KEY"

Then ask: "Generate a 1536x1024 hero image of a lighthouse at dusk in flat illustration style and save it to public/hero.png."

One flat $0.04 per image, for generations and edits, at any size and quality. The usage object in the response is informational and doesn't change the price.

When you submit an image, its price is reserved from your balance and settled when the image is done. You are not charged when a request is refused, a job is cancelled or no image is produced. Once an image has been generated it is charged, even if your client disconnected before it arrived. There is no subscription; see pricing for the other products on the same balance.