Text to Image

Generate an image from a text prompt with Qwen Image or Qwen Image 2.1. This endpoint creates an asynchronous task; use the returned task_id with Get task to retrieve the result.

Models and pricing

Model	Task type	Price per image	Default steps	Maximum steps
`Qubico/qwen-image-2.1`	`txt2img`	$0.015	8	16
`Qubico/qwen-image`	`txt2img`	$0.015	8	16

Select the matching model variant in Body Params. Existing Qwen Image clients can use 2.1 by changing model to Qubico/qwen-image-2.1; review any explicitly supplied steps and dimensions against the limits below.

Qwen Image 2.1 parameters

Qwen Image 2.1 uses the Turbo model for every request, with 1–16 steps (default and recommended: 8). Eight steps retain the official evaluated sampling schedule. Steps 1–7 keep the existing service-defined shorter schedule. Steps 9–16 use a service-defined schedule with deterministic piecewise linear interpolation over the nine official sigma anchors, including the final zero. These custom schedules are not an officially evaluated recipe and may reduce image quality; 9–16 steps are not guaranteed to outperform the recommended 8 steps.

prompt is required. negative_prompt is accepted for request compatibility but is not applied by Turbo at any supported step count (no classifier-free guidance).

width and height default to 1024. Each dimension must be from 64 to 4096 pixels, and width * height must not exceed 1,763,584 pixels (1328 × 1328). Both constraints apply: 4096 × 64 is valid, but 4096 × 4096 is not. Invalid dimensions are rejected before credits are reserved.

steps defaults to 8 when omitted or non-positive. Use 1–16; values above 16 are capped at 16.

seed uses a random seed when omitted or non-positive. Supply a positive integer to select a seed.

flow_shift and negative_prompt are accepted for request compatibility but ignored for all 2.1 requests (Turbo uses no classifier-free guidance).

Each request generates one image; denoise is fixed at 1.

Qwen Image 2.1 supports text-to-image only. For image editing, use Qubico/qwen-image on the Image Edit endpoint.

Existing Qwen Image

Qubico/qwen-image remains available. It keeps its default of 8 steps and maximum of 16, with an output-area limit of 1,048,576 pixels (1024 × 1024). Each dimension is 64–4096 pixels (default 1024), subject to that total-area limit: 4096 × 64 is valid, while 4096 × 4096 is not. Each model has its own parameter variant below. The legacy model also honors flow_shift values from 1 to 7; omission or null defaults to 1, values below 1 become 1, and values above 7 become 7. Set 3 explicitly if desired. negative_prompt: Accepted and encoded, but the current legacy workflow uses CFG=1, so negative_prompt does not provide classifier-free negative guidance.

Task usage

One generated image costs 150000 points ($0.015). A newly accepted task can show these points as meta.usage.frozen; after successful completion they appear in meta.usage.consume, with no remaining frozen points for that task. An accepted pending-task cancellation releases the reserved credits without a generation charge. Omitted numeric inputs can be echoed as 0 in the task response; this does not change the generation defaults above.

At 1 step, some outputs can be heavily blurred and lose readable text. Use the default 8 steps for general use.

Request

Header Params

Body Params application/jsonRequired

Examples

Qwen Image 2.1Qwen Image (existing model)

{
  "model": "Qubico/qwen-image-2.1",
  "task_type": "txt2img",
  "input": {
    "prompt": "A small ceramic teapot beside a window, clean product photograph, no text",
    "width": 1024,
    "height": 1024,
    "steps": 8,
    "seed": -1
  }
}

Request Code Samples

Shell

JavaScript

Java

Swift

PHP

Python

HTTP

Objective-C

Ruby

OCaml

Dart

Responses

🟢200Success

application/json

Bodyapplication/json

Examples

Qwen Image 2.1 pendingQwen Image (existing model) pending

Illustrative response; task ID and output URL are placeholders.

{
  "code": 200,
  "data": {
    "task_id": "00000000-0000-4000-8000-000000000001",
    "model": "Qubico/qwen-image-2.1",
    "task_type": "txt2img",
    "status": "pending",
    "config": {
      "service_mode": "",
      "webhook_config": {
        "endpoint": "",
        "secret": ""
      }
    },
    "input": {
      "prompt": "A small ceramic teapot beside a window, clean product photograph, no text",
      "width": 1024,
      "height": 1024,
      "steps": 8,
      "seed": -1,
      "negative_prompt": ""
    },
    "output": null,
    "meta": {
      "created_at": "2026-10-12T05:31:13.737522909Z",
      "started_at": "0001-01-01T00:00:00Z",
      "ended_at": "0001-01-01T00:00:00Z",
      "usage": {
        "type": "point",
        "frozen": 150000,
        "consume": 0
      },
      "is_using_private_pool": false
    },
    "detail": null,
    "logs": [],
    "error": {
      "code": 0,
      "message": "",
      "detail": null
    }
  },
  "message": "success"
}

Text to Image

Models and pricing#

Qwen Image 2.1 parameters#

Existing Qwen Image#