API documentation

Authentication

Create a key under API keys and send it in the Authorization header. Keys are shown once; store them as secrets and never ship them in a public web page.

Authorization: Bearer sk-...

Quick start

By default the call waits and returns the finished image. A typical image takes 5 to 40 seconds.

JavaScript

const res = await fetch("https://api.xelentapi.com/v1/api/generate", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.API_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    model: "nano-banana-2",
    prompt: "a paper boat on a puddle, overcast light",
    aspectRatio: "16:9",
    imageSize: "2K",
  }),
});
const job = await res.json();
console.log(job.status, job.results?.[0]?.url);

Python

import os, requests

r = requests.post(
    "https://api.xelentapi.com/v1/api/generate",
    headers={"Authorization": f"Bearer {os.environ['API_KEY']}"},
    json={
        "model": "gpt-image-2",
        "prompt": "a vintage travel poster of Lisbon",
        "aspectRatio": "2:3",
    },
    timeout=600,
)
job = r.json()
print(job["status"], job.get("results", [{}])[0].get("url"))

Generate

POST /v1/api/generate creates one image or one video.

A successful response:

{
  "id": "gen_7QmXv2...",
  "status": "succeeded",
  "progress": 100,
  "model": "nano-banana-2",
  "results": [{ "url": "https://api.xelentapi.com/files/gen_7QmXv2.../0.png", "type": "image/png" }],
  "cost_usd": 0.016,
  "created_at": 1790000000
}

status is one of running, succeeded, failed or violation (blocked by content moderation). Only succeeded is charged.

Parameters

Every model

Nano Banana models

GPT Image models

MiniMax H3 video

Reply types

json waits and returns the finished result. If a job runs past 6 minutes (20 for video) you get it back with status: "running"; poll it with its id.

stream returns text/event-stream with one event per progress change. The last event is the finished result.

data: {"id":"gen_...","status":"running","progress":0}

data: {"id":"gen_...","status":"running","progress":30}

data: {"id":"gen_...","status":"succeeded","progress":100,"results":[{"url":"https://api.xelentapi.com/files/gen_.../0.png","type":"image/png"}],"cost_usd":0.016}

async returns the id straight away. Use it for video and for batches.

# 1. submit
curl https://api.xelentapi.com/v1/api/generate -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"minimax-h3","prompt":"waves at dusk","resolution":"768p","duration":6,"replyType":"async"}'
# → {"id":"gen_...","status":"running"}

# 2. poll every few seconds
curl "https://api.xelentapi.com/v1/api/result?id=gen_..." -H "Authorization: Bearer $API_KEY"
# → {"id":"gen_...","status":"running","progress":40}
# → {"id":"gen_...","status":"succeeded","results":[{"url":"https://api.xelentapi.com/files/gen_.../0.mp4","type":"video/mp4"}],"cost_usd":0.12}

Get a result

GET /v1/api/result?id=gen_... (or POST with {"id": "..."}) returns the current state of a generation made with your account. Results stay available for as long as your files do.

OpenAI-compatible images

POST /v1/images/generations accepts the OpenAI request shape: model, prompt, size, quality, background, response_format (url or b64_json). n must be 1. For Nano Banana models, size is mapped to the nearest aspect ratio.

from openai import OpenAI

client = OpenAI(base_url="https://api.xelentapi.com/v1", api_key=os.environ["API_KEY"])
img = client.images.generate(model="gpt-image-2", prompt="an isometric tiny kitchen", size="1024x1024")
print(img.data[0].url)

Models and balance

GET /v1/models lists every model with its price. GET /v1/account returns your balance and the key's spend.

{"balance_usd": 12.4, "held_usd": 0.05, "key": {"name": "prod", "spent_usd": 3.2, "spend_limit_usd": 50}}

Errors

Errors return a non-2xx status with {"status": "failed", "error": "message", "code": "machine_code"}. You are never charged for an error.

Limits and files

• 120 requests per minute per key.
• 10 jobs running at once per account by default. Contact us to raise it.
• Request bodies up to 40 MB. Send large reference images as URLs.
• Output files are kept for 7 days. Download anything you want to keep.