Qwen Try-On API

Image generation and editing with Qwen-Image-2.1. JSON in, JSON out. No API key required.

Base URL http://ownv2.neuraa.in

Endpoints

Method	Path	Purpose
POST	`/tryon`	Generate / edit an image
GET	`/tryon/{jobId}`	Status or result of an async job
GET	`/jobs/{jobId}`	Saved job: prompt, params, input and output images
GET	`/health`	Service and model readiness
GET	`/openapi.json`	OpenAPI 3 spec
GET	`/swagger`	Interactive Swagger UI (needs internet for its assets)

POST/tryon

Send a prompt, optional negative prompt, optional images and parameters. Waits for the result and returns the image as base64 PNG.

Request body (JSON)

Field	Type	Required	Description
`prompt`	string	yes	What to generate or change (max 8000 chars). Refer to inputs as image 1, image 2 ...
`negative_prompt`	string	no	What to avoid. Needs `true_cfg_scale` above 1.
`images`	array of string	no	0-10 images: https URL, data URL or raw base64 (max 25 MB each).
`params`	object	no	Settings below. May also be sent at the top level.
`async`	boolean	no	true = return a jobId at once (202), then poll GET /tryon/{jobId}.

params

Param	Range	Default	Description
`seed`	integer	random	Same seed + inputs = same image.
`steps`	1 - 100	30	More steps = slower, finer.
`true_cfg_scale`	1 - 20	1	Guidance (alias `guidance_scale`). Use 3-5 so the negative prompt works.
`width` / `height`	256 - 2752	auto	Send both together. Rounded down to a multiple of 16.
`output_resolution`	512 - 2048	1024	Used when width/height are not set.
`num_images`	1 - 4	1	Number of variations returned.
`rgba`	true/false	false	Transparent background.

Example request

curl -X POST http://ownv2.neuraa.in/tryon \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Change the background to a sunset beach", "negative_prompt": "blurry, low quality", "images": ["https://example.com/photo.jpg", "data:image/png;base64,iVBORw0KGgo..."], "params": {"seed": 42, "steps": 30, "true_cfg_scale": 4.0, "width": 1024, "height": 1024, "num_images": 1}}'

{
  "prompt": "Change the background to a sunset beach",
  "negative_prompt": "blurry, low quality",
  "images": [
    "https://example.com/photo.jpg",
    "data:image/png;base64,iVBORw0KGgo..."
  ],
  "params": {
    "seed": 42,
    "steps": 30,
    "true_cfg_scale": 4.0,
    "width": 1024,
    "height": 1024,
    "num_images": 1
  }
}

Responses

200 Success - images are base64 PNGs

{
  "code": 200,
  "message": "ok",
  "data": {
    "jobId": "c463b6c90eac",
    "mimeType": "image/png",
    "seed": 42,
    "genSeconds": 8.2,
    "images": [
      "<base64 PNG>"
    ]
  }
}

202 Queued (only when async is true)

{
  "code": 202,
  "message": "queued",
  "data": {
    "jobId": "864da964a5c4",
    "state": "queued"
  }
}

400 Invalid input

{
  "code": 400,
  "message": "prompt required"
}

503 Model busy or loading

{
  "code": 503,
  "message": "model busy",
  "retryable": true
}

504 Timed out

{
  "code": 504,
  "message": "generation timed out",
  "jobId": "864da964a5c4",
  "retryable": true
}

Decode data.images[0] from base64 and save it as .png to get the picture.

GET/tryon/{jobId}

Poll every 1-3 s after sending "async": true. data.state is queued, running or done.

200 Running

{
  "code": 200,
  "message": "running",
  "data": {
    "jobId": "864da964a5c4",
    "state": "running",
    "step": 12,
    "steps": 30,
    "position": 0,
    "etaSeconds": 6.1
  }
}

200 Done

{
  "code": 200,
  "message": "ok",
  "data": {
    "jobId": "864da964a5c4",
    "state": "done",
    "mimeType": "image/png",
    "seed": 1,
    "genSeconds": 4.4,
    "images": [
      "<base64 PNG>"
    ]
  }
}

404 Unknown job

{
  "code": 404,
  "message": "job not found"
}

502 Generation failed

{
  "code": 502,
  "message": "generation error",
  "data": {
    "jobId": "864da964a5c4",
    "state": "error",
    "detail": "..."
  }
}

GET/health

200 Ready

{
  "status": "ok",
  "model": "ready"
}

503 Model loading or unreachable

{
  "status": "degraded",
  "model": "loading model..."
}

Saved jobs

Every job is saved with its job ID, prompt, negative prompt, parameters, input images and output images. The job ID is returned by POST /tryon (data.jobId, plus data.jobUrl).

Method	Path	Purpose
GET	`/jobs/{jobId}`	Full record of one job (open, by ID)
GET	`/jobs/{jobId}/input/{n}`	Input image n (PNG), n starts at 0
GET	`/jobs/{jobId}/output/{n}`	Output image n (PNG), n starts at 0
GET	`/jobs?limit=50&offset=0&status=done&q=text`	List all jobs, newest first. Needs the history key: `Authorization: Bearer <key>`

GET /jobs/{jobId}

200 Job record

{
  "code": 200,
  "message": "ok",
  "data": {
    "jobId": "303b04234bef",
    "status": "done",
    "prompt": "In image 1, change the top to a red t-shirt",
    "negativePrompt": "blurry",
    "createdAt": "2026-10-20T15:41:39+00:00",
    "genSeconds": 16.2,
    "totalSeconds": 16.6,
    "seed": 2126666741,
    "width": 1024,
    "height": 1024,
    "error": null,
    "params": {
      "steps": 20,
      "true_cfg_scale": 1.0,
      "num_images": 1
    },
    "inputImages": [
      {
        "index": 0,
        "url": "http://ownv2.neuraa.in/jobs/303b04234bef/input/0"
      }
    ],
    "outputImages": [
      {
        "index": 0,
        "url": "http://ownv2.neuraa.in/jobs/303b04234bef/output/0"
      }
    ]
  }
}

404 Unknown job

{
  "code": 404,
  "message": "job not found"
}

401 Listing without the key

{
  "code": 401,
  "message": "listing jobs needs the history key"
}

Errors

Every error is JSON: {"code": <http status>, "message": "..."}

HTTP	Meaning
400	Bad input (message names the problem: prompt required, parameter out of range, bad image/URL ...)
404	Unknown path or jobId
405	Wrong HTTP method
409	Job was cancelled
413	Request body too large (max 256 MB)
502	Generation / model server error
503	Model busy or still loading (retryable)
504	Timed out after 600 s (use async)

Image input

Each item in images is an https URL, a data URL (data:image/png;base64,...) or raw base64. Max 10 images, 25 MB each. URLs must be https and public (private/localhost addresses are rejected). If a site blocks downloads, send the image as base64.