Images

Image models use OpenAI's Images API, so images.generate in any OpenAI SDK works with the same base URL and key as text.

import { writeFile } from "node:fs/promises"
import OpenAI from "openai"

const ai = new OpenAI({
  apiKey: process.env.UPSKY_API_KEY,
  baseURL: "https://api.upsky.dev/v1/ai",
})

const image = await ai.images.generate({
  model: "bfl/flux-2-pro",
  prompt: "A lighthouse on a cliff at dawn, watercolor",
  size: "1024x1024",
})

await writeFile("lighthouse.png", Buffer.from(image.data[0].b64_json, "base64"))

Get the image models from the models endpoint with ?type=image.

OptionsCopy Anchor Link

• n: how many images to make, from 1 to 4. Each one is charged.
• size: width and height in pixels, such as 1024x1024. Each model supports its own sizes.
• aspect_ratio: width to height, such as 16:9, for models that take a ratio instead of a size.
• seed: makes results repeatable, on models that support it.

Images always come back as base64 in b64_json, never as links, and Upsky doesn't keep them. Other fields, such as quality or style, are ignored.

What it costsCopy Anchor Link

Most image models charge a set price per image, listed on the models endpoint as pricing.image. Some charge more for larger sizes, listed under pricing.image_sizes. A size between two listed ones costs what the next one up does, and a size larger than the largest is refused.

A few models, such as OpenAI's, are priced by tokens, like a text model. Upsky sets aside an estimate before the request and charges what the provider reports.

You're charged only for the images that come back. usage.cost on the reply says what the request cost, in US dollars.

Test keysCopy Anchor Link

A test key gets a 1×1 transparent PNG for each image asked for. It never calls a model and is never charged.