# GPT Image 2 API via PiAPI

Base URL: https://api.piapi.ai
Auth: X-API-KEY header (get your key at https://piapi.ai/workspace/key)

Documentation: https://piapi.ai/docs

## Available Models

Token-billed models charge by output tokens; per-image costs below are
estimates for 1024x1024 at the given quality.

- gpt-image-2: $30 / 1M output tokens (~$0.032/image medium). Currently only accepts quality = "medium".
- gpt-image-2-preview: $0.10 flat per image
- gpt-image-1.5: $32 / 1M output tokens (low ~$0.009 / medium ~$0.034 / high ~$0.133 per image)
- gpt-image-1: $40 / 1M output tokens (low ~$0.011 / medium ~$0.042 / high ~$0.167 per image)

## Create Task (async)

POST /api/v1/images/generations/async
{
  "model": "gpt-image-2",
  "prompt": "A luxury serum bottle above reflective glass, cyan rim lighting",
  "size": "1024x1024",
  "quality": "medium"
}

Note: gpt-image-2 currently only supports quality = "medium".

## Get Task Result

GET /api/v1/task/{task_id}

Status: "pending" → "processing" → "completed" or "failed"
Image: response.data.output.data[0].b64_json (base64-encoded PNG)
Billing: response.data.output.usage.output_tokens × the model's rate
  (gpt-image-2 = $30 / 1M output tokens)

## Quick Start

1. Get API key: https://piapi.ai/workspace/key
2. Top up: https://piapi.ai/workspace/billing
3. POST /api/v1/images/generations/async
4. Poll GET /api/v1/task/{task_id} until "completed"
5. Decode the image from response.data.output.data[0].b64_json