Endpoints

Image generation

Generate an image from a text prompt. OpenAI-compatible images endpoint.

POST /api/v1/images/generations

curl https://api.pendra.ai/api/v1/images/generations \
  -H "Authorization: Bearer pdr_sk_..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "flux.1-schnell",
    "prompt": "A misty Welsh valley at dawn, painted in oils"
  }'

from pendra import Pendra
import base64, pathlib

client = Pendra()
result = client.images.generations.create(
    model="flux.1-schnell",
    prompt="A misty Welsh valley at dawn, painted in oils",
)
pathlib.Path("out.png").write_bytes(base64.b64decode(result.data[0].b64_json))

import Pendra from 'pendra';
import { writeFileSync } from 'node:fs';

const client = new Pendra();
const result = await client.images.generations.create({
  model: 'flux.1-schnell',
  prompt: 'A misty Welsh valley at dawn, painted in oils',
});
writeFileSync('out.png', Buffer.from(result.data[0].b64_json, 'base64'));

200

{
  "created": 1715346200,
  "data": [
    { "b64_json": "iVBORw0KGgoAAAANSUhEUgAA...Cg==" }
  ]
}

New to image generation? Start with the Image generation guide. This page is the field-by-field reference.

Body application/json

model string required

Image model ID, e.g. sdxl-turbo. List models at /models?type=image.

prompt string required

The text description of the image to generate.

size string default: model's native resolution

Output dimensions as WxH. When omitted, Pendra uses the model's native resolution (512x512 for stable-diffusion-1.5 and sdxl-turbo; 1024x1024 for sdxl-base-1.0, flux.1-schnell, and stable-diffusion-3.5-large-turbo), which gives the best quality. Supported sizes depend on the model.

n integer default: 1

Number of images to generate (up to 4).

negative_prompt string

Text describing what to avoid in the image.

num_inference_steps integer default: model-dependent

Diffusion steps — higher is slower but can be sharper. When omitted, defaults per model: ~30 for standard checkpoints (stable-diffusion-1.5, sdxl-base-1.0) and 4 for distilled turbo/schnell models, which converge in a handful of steps.

seed integer

Fix the seed for reproducible output.

response_format string default: b64_json

Only b64_json is supported — responses always return a base64-encoded PNG.

Response

An OpenAI-shaped image response. data contains one entry per requested image, each carrying a base64-encoded PNG in b64_json — decode it before writing to disk.

Timeouts

Image requests can run up to ~30 minutes, since generation can be slow on busy workers. Pendra keeps the connection alive while the worker renders, so a long job won't be cut off mid-generation. For interactive UX, render a spinner; you'll usually get a response in 5–30 seconds depending on model and size.

Usage tracking

Image requests appear under Images in the console usage view alongside chat and embeddings.