---
title: Qwen-Image-2512 | Runware Docs
url: https://runware.ai/docs/models/alibaba-qwen-image-2512
description: Advanced multimodal image generation model
---
# Qwen-Image-2512

![Qwen-Image-2512](https://assets.runware.ai/covers/alibaba-qwen-image-2512.jpg)

Qwen-Image-2512 is an improved version of the Qwen-Image image foundation model with enhanced prompt understanding, superior text rendering accuracy, and more realistic visual details. It generates high-fidelity images from text prompts across diverse subjects and styles.

- **Status**: live
- **ID**: `alibaba:qwen-image@2512`
- Try in Playground
- **Creator**: Alibaba
- **Release Date**: December 25, 2025
- **Zero Data Retention**: Yes
- **Capabilities**: Text to Image, Image to Image, Checkpoint

## Pricing

$0.00364 from, per run · ~11s per request

**Measured runs**

- 512×512 · steps 20: `$0.00364` (~4s)
- 1024×1024 · steps 20: `$0.012` (~10s)
- 1024×1024 · steps 40: `$0.0227` (~20s)
- 2048×2048 · steps 20: `$0.0598` (~53s)

**4 example generations**

- [image-to-image](https://runware.ai/docs/models/alibaba-qwen-image-2512/examples#example-raptor-center-annual-report-portrait) · 1024×1536: `$0.0032` (~8s)
- [expand](https://runware.ai/docs/models/alibaba-qwen-image-2512/examples#example-refrigerated-rail-campaign-key-visual) · 1536×1024: `$0.0045` (~9s)
- [inpaint](https://runware.ai/docs/models/alibaba-qwen-image-2512/examples#example-vaccine-cooler-field-lifestyle-shot) · 1280×1280: `$0.0058` (~12s)
- [text-to-image](https://runware.ai/docs/models/alibaba-qwen-image-2512/examples#example-pipe-organ-crew-portrait) · 1536×1024: `$0.0064` (~11s)

billed by compute time, so a request costs what it took · times vary, measured end to end · [How pricing works](https://runware.ai/docs/models-api/pricing)

## Compatibility & Validation

When `strength` or `inputs.maskImage` is provided, `inputs.seedImage` is required.

---

`width` and `height` must be used together.

## Request Parameters

**API Options**

Platform-level options for task execution and delivery.

### [taskType](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-tasktype)

- **Type**: `string`
- **Required**: true
- **Value**: `imageInference`

Identifier for the type of task being performed.

### [taskUUID](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-taskuuid)

- **Type**: `string`
- **Required**: true
- **Format**: `UUID v4`

UUID v4 identifier for tracking tasks and matching async responses. Must be unique per task.

### [outputType](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-outputtype)

- **Type**: `string`
- **Default**: `URL`

Image output type.

**Allowed values**: `URL` `base64Data` `dataURI`

### [outputFormat](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-outputformat)

- **Type**: `string`
- **Default**: `JPG`

Specifies the file format of the generated output. The available values depend on the task type and the specific model's capabilities.

- \`JPG\`: Best for photorealistic images with smaller file sizes (no transparency).
- \`PNG\`: Lossless compression, supports high quality and transparency (alpha channel).
- \`WEBP\`: Modern format providing superior compression and transparency support.

> [!NOTE]
> \*\*Transparency\*\*: If you are using features like background removal or LayerDiffuse that require transparency, you must select a format that supports an alpha channel (e.g., \`PNG\`, \`WEBP\`, \`TIFF\`). \`JPG\` does not support transparency.

**Allowed values**: `JPG` `PNG` `WEBP`

### [outputQuality](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-outputquality)

- **Type**: `integer`
- **Min**: `20`
- **Max**: `99`
- **Default**: `95`

Compression quality of the output. Higher values preserve quality but increase file size.

### [webhookURL](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-webhookurl)

- **Type**: `string`
- **Format**: `uri`

Webhook URL that receives JSON responses via HTTP POST when generation tasks complete. For batch requests with multiple results, each completed item triggers a separate webhook call as it becomes available.

**Learn more** (1 resource):

- [Webhooks](https://runware.ai/docs/models-api/webhooks) (platform)

### [deliveryMethod](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-deliverymethod)

- **Type**: `string`
- **Default**: `sync`

Determines how the API delivers task results.

**Allowed values**:

- `sync` Returns results directly in the API response. Over HTTP, a task still running after 120 seconds continues as an async task and comes back with status 'processing'. Poll for its results using getResponse.
- `async` Returns an immediate acknowledgment with the task UUID. Poll for results using getResponse.

**Learn more** (1 resource):

- [Task Polling](https://runware.ai/docs/models-api/task-polling) (platform)

### [uploadEndpoint](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-uploadendpoint)

- **Type**: `string`
- **Format**: `uri`

Specifies a URL where the generated content will be automatically uploaded using the HTTP PUT method. The raw binary data of the media file is sent directly as the request body. For secure uploads to cloud storage, use presigned URLs that include temporary authentication credentials.

**Common use cases:**

- **Cloud storage**: Upload directly to S3 buckets, Google Cloud Storage, or Azure Blob Storage using presigned URLs.
- **CDN integration**: Upload to content delivery networks for immediate distribution.

```text
// S3 presigned URL for secure upload
https://your-bucket.s3.amazonaws.com/generated/content.mp4?X-Amz-Signature=abc123&X-Amz-Expires=3600

// Google Cloud Storage presigned URL
https://storage.googleapis.com/your-bucket/content.jpg?X-Goog-Signature=xyz789

// Custom storage endpoint
https://storage.example.com/uploads/generated-image.jpg
```

The content data will be sent as the request body to the specified URL when generation is complete.

### [safety](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-safety)

- **Path**: `safety.checkContent`
- **Type**: `object (1 property)`

Content safety checking configuration for image generation.

#### [checkContent](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-safety-checkcontent)

- **Path**: `safety.checkContent`
- **Type**: `boolean`

Enable or disable content safety checking. Increases total generation time and adds a small charge on top of the generation cost.

### [ttl](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-ttl)

- **Type**: `integer`
- **Min**: `60`

Time-to-live (TTL) in seconds for generated content. Only applies when `outputType` is `URL`.

### [includeCost](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-includecost)

- **Type**: `boolean`

Include task cost in the response.

### [numberResults](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-numberresults)

- **Type**: `integer`
- **Min**: `1`
- **Max**: `20`
- **Default**: `1`

Number of results to generate. Each result uses a different seed, producing variations of the same parameters.

### [acceleratorOptions](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-acceleratoroptions)

- **Path**: `acceleratorOptions.cacheEndStep`
- **Type**: `object (12 properties)`

Advanced caching mechanisms to speed up generation.

#### [cacheEndStep](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-acceleratoroptions-cacheendstep)

- **Path**: `acceleratorOptions.cacheEndStep`
- **Type**: `integer`
- **Min**: `1`

Absolute step number to end caching. Must be greater than `cacheStartStep` and less than or equal to `steps`.

#### [cacheEndStepPercentage](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-acceleratoroptions-cacheendsteppercentage)

- **Path**: `acceleratorOptions.cacheEndStepPercentage`
- **Type**: `integer`
- **Min**: `1`
- **Max**: `100`

Percentage of steps to end caching. Alternative to `cacheEndStep`. Must be greater than `cacheStartStepPercentage`.

#### [cacheMaxConsecutiveSteps](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-acceleratoroptions-cachemaxconsecutivesteps)

- **Path**: `acceleratorOptions.cacheMaxConsecutiveSteps`
- **Type**: `integer`
- **Min**: `1`
- **Max**: `5`
- **Default**: `3`

Limits the maximum number of consecutive steps that can use cached computations before forcing a fresh computation.

#### [cacheStartStep](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-acceleratoroptions-cachestartstep)

- **Path**: `acceleratorOptions.cacheStartStep`
- **Type**: `integer`
- **Min**: `0`

Absolute step number to start caching. Must be less than `cacheEndStep`.

#### [cacheStartStepPercentage](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-acceleratoroptions-cachestartsteppercentage)

- **Path**: `acceleratorOptions.cacheStartStepPercentage`
- **Type**: `integer`
- **Min**: `0`
- **Max**: `99`

Percentage of steps to start caching. Alternative to `cacheStartStep`. Must be less than `cacheEndStepPercentage`.

#### [fbCache](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-acceleratoroptions-fbcache)

- **Path**: `acceleratorOptions.fbCache`
- **Type**: `boolean`

First Block Cache (FBCache) acceleration. Reuses feature block computations across steps.

#### [fbCacheThreshold](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-acceleratoroptions-fbcachethreshold)

- **Path**: `acceleratorOptions.fbCacheThreshold`
- **Type**: `float`
- **Min**: `0`
- **Max**: `1`
- **Step**: `0.01`
- **Default**: `0.25`

Controls the sensitivity threshold for determining when to reuse cached computations. Lower values reuse more aggressively.

#### [teaCache](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-acceleratoroptions-teacache)

- **Path**: `acceleratorOptions.teaCache`
- **Type**: `boolean`

TeaCache acceleration for transformer-based models. Estimates step differences to skip redundant computations.

#### [teaCacheDistance](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-acceleratoroptions-teacachedistance)

- **Path**: `acceleratorOptions.teaCacheDistance`
- **Type**: `float`
- **Min**: `0`
- **Max**: `1`
- **Step**: `0.01`
- **Default**: `0.5`

Controls the aggressiveness of the TeaCache feature. Lower values prioritize quality, higher values prioritize speed.

#### [dbCache](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-acceleratoroptions-dbcache)

- **Path**: `acceleratorOptions.dbCache`
- **Type**: `boolean`

DB Cache (CacheDiT) acceleration. Caches and reuses intermediate transformer block outputs to skip redundant computations.

#### [dbCacheThreshold](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-acceleratoroptions-dbcachethreshold)

- **Path**: `acceleratorOptions.dbCacheThreshold`
- **Type**: `float`
- **Min**: `0`
- **Max**: `1`
- **Step**: `0.01`
- **Default**: `0.25`

Controls the sensitivity threshold for DB Cache. Lower values reuse cached blocks more aggressively, higher values prioritize quality.

#### [dbCacheSkipInterval](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-acceleratoroptions-dbcacheskipinterval)

- **Path**: `acceleratorOptions.dbCacheSkipInterval`
- **Type**: `integer`
- **Min**: `1`
- **Default**: `5`

Controls how many steps to skip between cache refreshes. Higher values skip more steps for faster generation at the cost of quality.

**Inputs**

Input resources for the task (images, audio, etc). These must be nested inside the \`inputs\` object.

### [seedImage](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-inputs-seedimage)

- **Path**: `inputs.seedImage`
- **Type**: `string`

Image used as a starting point for the generation (UUID, URL, Data URI, or Base64).

**Learn more** (3 resources):

- [Image To Image: Seed Image The Foundation](https://runware.ai/docs/learn/image-to-image#seed-image-the-foundation) (learn)
- [Image Inpainting: Seed And Mask Image The Foundation](https://runware.ai/docs/learn/image-inpainting#seed-and-mask-image-the-foundation) (learn)
- [Image Outpainting: Seed Image The Starting Point](https://runware.ai/docs/learn/image-outpainting#seed-image-the-starting-point) (learn)

### [maskImage](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-inputs-maskimage)

- **Path**: `inputs.maskImage`
- **Type**: `string`

Image used to specify which areas of the input image should be edited (UUID, URL, Data URI, or Base64). White marks the area to edit, black is preserved.

**Learn more** (1 resource):

- [Image Inpainting: Seed And Mask Image The Foundation](https://runware.ai/docs/learn/image-inpainting#seed-and-mask-image-the-foundation) (learn)

**Core Parameters**

Primary parameters that define the task output.

### [model](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-model)

- **Type**: `string`
- **Required**: true
- **Value**: `alibaba:qwen-image@2512`

Send `auto` instead of an AIR identifier to have the model router choose the model for this task. The rest of the request is unchanged, every value you send is kept, and only the parameters you leave out are filled in. See [Model Router](https://runware.ai/docs/models-api/model-router).

**Learn more** (3 resources):

- [Text To Image: Model Selection](https://runware.ai/docs/learn/text-to-image#model-selection) (learn)
- [Image Inpainting: Model Specialized Inpainting Models](https://runware.ai/docs/learn/image-inpainting#model-specialized-inpainting-models) (learn)
- [Image Outpainting: Other Critical Parameters](https://runware.ai/docs/learn/image-outpainting#other-critical-parameters) (learn)

### [positivePrompt](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-positiveprompt)

- **Type**: `string`
- **Required**: true
- **Min**: `2`
- **Max**: `32000`

Text prompt describing elements to include in the generated output.

**Learn more** (1 resource):

- [Prompts](https://runware.ai/docs/learn/prompts) (learn)

### [negativePrompt](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-negativeprompt)

- **Type**: `string`
- **Min**: `2`
- **Max**: `3000`

Prompt to guide what to exclude from generation. Ignored when guidance is disabled (CFGScale ≤ 1).

**Learn more** (1 resource):

- [Prompts: Negative Prompts](https://runware.ai/docs/learn/prompts#negative-prompts) (learn)

### [width](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-width)

- **Type**: `integer`
- **Required**: true
- **Min**: `256`
- **Max**: `2048`
- **Step**: `16`

Width of the generated media in pixels.

**Learn more** (2 resources):

- [Dimensions](https://runware.ai/docs/learn/dimensions) (learn)
- [Image Outpainting: Dimensions Critical For Outpainting](https://runware.ai/docs/learn/image-outpainting#dimensions-critical-for-outpainting) (learn)

### [height](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-height)

- **Type**: `integer`
- **Required**: true
- **Min**: `256`
- **Max**: `2048`
- **Step**: `16`

Height of the generated media in pixels.

**Learn more** (2 resources):

- [Dimensions](https://runware.ai/docs/learn/dimensions) (learn)
- [Image Outpainting: Dimensions Critical For Outpainting](https://runware.ai/docs/learn/image-outpainting#dimensions-critical-for-outpainting) (learn)

### [seed](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-seed)

- **Type**: `integer`
- **Min**: `0`
- **Max**: `9223372036854776000`

Random seed for reproducible generation. When not provided, a random seed is generated in the unsigned 32-bit range.

**Learn more** (1 resource):

- [Seed](https://runware.ai/docs/learn/seed) (learn)

### [steps](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-steps)

- **Type**: `integer`
- **Min**: `1`
- **Max**: `100`
- **Default**: `20`

Total number of denoising steps. Higher values generally produce more detailed results but take longer.

**Learn more** (1 resource):

- [Steps](https://runware.ai/docs/learn/steps) (learn)

### [scheduler](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-scheduler)

- **Type**: `string`

Scheduler to use for the diffusion process.

**Allowed values**: `DDIM` `DDIMScheduler` `DDPMScheduler` `DEISMultistepScheduler` `Default` `DPM++` `DPM++ 2M` `DPM++ 2M Beta` `DPM++ 2M Exponential` `DPM++ 2M Karras` `DPM++ 2M SDE` `DPM++ 2M SDE Beta` `DPM++ 2M SDE Exponential` `DPM++ 2M SDE Karras` `DPM++ 2M SDE Uniform` `DPM++ 2M Uniform` `DPM++ 3M` `DPM++ 3M Beta` `DPM++ 3M Exponential` `DPM++ 3M Karras` `DPM++ 3M SDE Uniform` `DPM++ 3M Uniform` `DPM++ Beta` `DPM++ Exponential` `DPM++ Karras` `DPM++ SDE` `DPM++ SDE Beta` `DPM++ SDE Exponential` `DPM++ SDE Karras` `DPM++ Uniform` `DPM++ Uniform Beta` `DPM++ Uniform Exponential` `DPM++ Uniform Karras` `DPMSolverMultistepInverse` `DPMSolverMultistepScheduler` `DPMSolverSinglestepScheduler` `EDMDPMSolverMultistepScheduler` `EDMEulerScheduler` `Euler` `Euler a` `Euler Beta` `Euler Exponential` `Euler Karras` `EulerAncestralDiscreteScheduler` `EulerDiscreteScheduler` `FlowMatchEulerDiscreteScheduler` `Heun` `HeunDiscreteScheduler` `Heun Karras` `IPNDMScheduler` `IPNDM Uniform` `IPNDM Uniform Beta` `IPNDM Uniform Exponential` `IPNDM Uniform Karras` `KDPM2AncestralDiscreteScheduler` `KDPM2DiscreteScheduler` `LCM` `LCMScheduler` `LMS` `LMSDiscreteScheduler` `LMS Karras` `PNDMScheduler` `TCDScheduler` `UniPC` `UniPC 2M` `UniPC 2M Karras` `UniPC 2M Uniform` `UniPC 3M` `UniPC 3M Karras` `UniPC 3M Uniform` `UniPC Karras` `UniPC Uniform` `UniPC Uniform Beta` `UniPC Uniform Exponential` `UniPC Uniform Karras`

**Learn more** (2 resources):

- [Schedulers](https://runware.ai/docs/learn/schedulers) (learn)
- [Image Outpainting: Other Critical Parameters](https://runware.ai/docs/learn/image-outpainting#other-critical-parameters) (learn)

### [CFGScale](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-cfgscale)

- **Type**: `float`
- **Min**: `0`
- **Step**: `0.01`

Guidance scale representing how closely the output will resemble the prompt. Higher values produce results more aligned with the prompt.

**Learn more** (1 resource):

- [Cfg Scale](https://runware.ai/docs/learn/cfg-scale) (learn)

### [strength](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-strength)

- **Type**: `float`
- **Min**: `0`
- **Max**: `1`
- **Step**: `0.01`
- **Default**: `0.8`

Strength of the transformation. Lower values result in more influence from the original input.

**Learn more** (3 resources):

- [Image To Image: Strength The Transformation Intensity](https://runware.ai/docs/learn/image-to-image#strength-the-transformation-intensity) (learn)
- [Image Inpainting: Strength Controlling Transformation Intensity](https://runware.ai/docs/learn/image-inpainting#strength-controlling-transformation-intensity) (learn)
- [Image Outpainting: Strength Understanding The Critical Threshold](https://runware.ai/docs/learn/image-outpainting#strength-understanding-the-critical-threshold) (learn)

### [maskMargin](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-maskmargin)

- **Type**: `integer`
- **Min**: `32`
- **Max**: `128`

Extra context pixels around the masked region during inpainting. The model zooms into the masked area with these additional pixels for better integration.

**Learn more** (1 resource):

- [Image Inpainting: Mask Margin Enhancing Detail](https://runware.ai/docs/learn/image-inpainting#mask-margin-enhancing-detail) (learn)

### [outpaint](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-outpaint)

- **Path**: `outpaint.top`
- **Type**: `object (5 properties)`

Pixel extensions for each boundary direction of the source image. At least one direction is required.

**Learn more** (1 resource):

- [Image Outpainting: Outpaint Defining The Expansion](https://runware.ai/docs/learn/image-outpainting#outpaint-defining-the-expansion) (learn)

#### [top](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-outpaint-top)

- **Path**: `outpaint.top`
- **Type**: `integer`
- **Min**: `0`
- **Step**: `64`

Number of pixels to extend to the top.

#### [bottom](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-outpaint-bottom)

- **Path**: `outpaint.bottom`
- **Type**: `integer`
- **Min**: `0`
- **Step**: `64`

Number of pixels to extend to the bottom.

#### [left](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-outpaint-left)

- **Path**: `outpaint.left`
- **Type**: `integer`
- **Min**: `0`
- **Step**: `64`

Number of pixels to extend to the left.

#### [right](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-outpaint-right)

- **Path**: `outpaint.right`
- **Type**: `integer`
- **Min**: `0`
- **Step**: `64`

Number of pixels to extend to the right.

#### [blur](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-outpaint-blur)

- **Path**: `outpaint.blur`
- **Type**: `integer`
- **Min**: `0`
- **Max**: `32`

Amount of blur to apply to the mask edge.

**Learn more** (1 resource):

- [Image Outpainting: Outpaint Defining The Expansion](https://runware.ai/docs/learn/image-outpainting#outpaint-defining-the-expansion) (learn)

### [lora](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-lora)

- **Path**: `lora.model`
- **Type**: `array of objects (2 properties)`
- **Min items**: `1`

With LoRA (Low-Rank Adaptation), you can adapt a model to specific styles or features by emphasizing particular aspects of the data. This technique enhances the quality and relevance of generated content and can be especially useful when the output needs to adhere to a specific artistic style or follow particular guidelines.

Multiple LoRA models can be used simultaneously to achieve different adaptation goals.

**Examples**:

```json
"lora": [
  {
    "model": "<lora-model-air>",
    "weight": 0.8
  }
]
```

**Learn more** (1 resource):

- [Loras](https://runware.ai/docs/learn/loras) (learn)

#### [model](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-lora-model)

- **Path**: `lora.model`
- **Type**: `string`
- **Required**: true

LoRA model identifier.

#### [weight](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-lora-weight)

- **Path**: `lora.weight`
- **Type**: `float`
- **Min**: `-4`
- **Max**: `4`
- **Step**: `0.01`
- **Default**: `1`

Strength of the LoRA influence. A value of 0 means no influence. Higher values increase the influence, and negative values can be used to steer away from the LoRA's style.

**Settings**

Technical parameters to fine-tune the inference process. These must be nested inside the \`settings\` object.

### [trueCFGScale](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-settings-truecfgscale)

- **Path**: `settings.trueCFGScale`
- **Type**: `float`

True Classifier-Free Guidance scale. Higher values increase prompt adherence at the cost of quality.

**Features**

Standalone addons and post-processing features.

### [ultralytics](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-ultralytics)

- **Path**: `ultralytics.CFGScale`
- **Type**: `object (9 properties)`

Configuration object for Ultralytics face enhancement during generation. This feature uses face detection and inpainting to improve facial details in the same generation step, without requiring post-processing.

> [!NOTE]
> Face enhancement is available for Stable Diffusion 1.X, SDXL, and FLUX models. The system automatically detects faces and applies targeted refinement to improve quality while maintaining consistency with the overall generation.

#### [CFGScale](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-ultralytics-cfgscale)

- **Path**: `ultralytics.CFGScale`
- **Type**: `float`
- **Min**: `0`
- **Max**: `50`
- **Step**: `0.1`
- **Default**: `8`

Face refinement guidance scale.

#### [confidence](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-ultralytics-confidence)

- **Path**: `ultralytics.confidence`
- **Type**: `float`
- **Min**: `0`
- **Max**: `1`
- **Step**: `0.01`
- **Default**: `0.9`

Confidence threshold for detection.

#### [inpaintSize](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-ultralytics-inpaintsize)

- **Path**: `ultralytics.inpaintSize`
- **Type**: `integer`
- **Min**: `128`
- **Max**: `2048`
- **Default**: `1024`

Image size (in pixels) to use for each inpainting region. YOLO detects faces, crops the region, and scales it to this size before running diffusion. Set so most faces land in the 2–4× range of their original pixel size. Going beyond 8× may degrade identity resemblance.

#### [maskBlur](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-ultralytics-maskblur)

- **Path**: `ultralytics.maskBlur`
- **Type**: `integer`
- **Min**: `0`
- **Max**: `100`
- **Default**: `5`

Mask feathering amount. Higher values create softer transitions between the enhanced face region and surrounding areas.

#### [maskPadding](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-ultralytics-maskpadding)

- **Path**: `ultralytics.maskPadding`
- **Type**: `integer`
- **Min**: `0`
- **Max**: `20`
- **Default**: `5`

Padding around detected face in pixels. Expands the refinement area to include surrounding context like hair and neck.

#### [negativePrompt](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-ultralytics-negativeprompt)

- **Path**: `ultralytics.negativePrompt`
- **Type**: `string`

Negative prompt for detection.

#### [positivePrompt](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-ultralytics-positiveprompt)

- **Path**: `ultralytics.positivePrompt`
- **Type**: `string`

Positive prompt for detection.

#### [steps](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-ultralytics-steps)

- **Path**: `ultralytics.steps`
- **Type**: `integer`
- **Min**: `1`
- **Max**: `100`
- **Default**: `20`

Number of face refinement steps.

#### [strength](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-ultralytics-strength)

- **Path**: `ultralytics.strength`
- **Type**: `float`
- **Max**: `1`
- **Step**: `0.01`
- **Default**: `0.3`

Refinement strength. Lower values preserve more of the original, higher values allow more aggressive reconstruction.

### [watermark](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-advancedfeatures-watermark)

- **Path**: `advancedFeatures.watermark`
- **Type**: `object (6 properties)`

Configuration object for adding watermarks to generated videos. Watermarks can be applied using either text or image content with customizable positioning and appearance. You must provide either `text` or `image` content for the watermark, but not both.

**Text watermark**:

```json
"advancedFeatures": {
  "watermark": {
    "text": "© 2025 Company",
    "displayPosition": "bottom-right",
    "opacity": 0.6,
    "fontColor": "#ffffff",
    "bgColor": "#000000"
  }
}
```

**Image watermark**:

```json
"advancedFeatures": {
  "watermark": {
    "image": "c64351d5-4c59-42f7-95e1-eace013eddab",
    "displayPosition": "top-left",
    "opacity": 0.6
  }
}
```

**Tiled watermark**:

```json
"advancedFeatures": {
  "watermark": {
    "text": "PREVIEW",
    "tiled": true,
    "opacity": 0.4,
    "fontColor": "#cccccc"
  }
}
```

#### [text](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-advancedfeatures-watermark-text)

- **Path**: `advancedFeatures.watermark.text`
- **Type**: `string`
- **Min**: `2`
- **Max**: `32`

Watermark text.

#### [image](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-advancedfeatures-watermark-image)

- **Path**: `advancedFeatures.watermark.image`
- **Type**: `string`

Watermark image (UUID, URL, Data URI, or Base64).

#### [displayPosition](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-advancedfeatures-watermark-displayposition)

- **Path**: `advancedFeatures.watermark.displayPosition`
- **Type**: `string`

Watermark position.

**Allowed values**: `top-left` `top-center` `top-right` `center-left` `center-center` `center-right` `bottom-left` `bottom-center` `bottom-right` `tiled`

#### [opacity](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-advancedfeatures-watermark-opacity)

- **Path**: `advancedFeatures.watermark.opacity`
- **Type**: `float`
- **Min**: `0.1`
- **Max**: `1`
- **Step**: `0.01`

Watermark opacity.

#### [fontColor](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-advancedfeatures-watermark-fontcolor)

- **Path**: `advancedFeatures.watermark.fontColor`
- **Type**: `string`

Text color in hex format.

#### [bgColor](https://runware.ai/docs/models/alibaba-qwen-image-2512#request-advancedfeatures-watermark-bgcolor)

- **Path**: `advancedFeatures.watermark.bgColor`
- **Type**: `string`

Background color in hex format.

## Response Parameters

### [taskType](https://runware.ai/docs/models/alibaba-qwen-image-2512#response-tasktype)

- **Type**: `string`
- **Required**: true
- **Value**: `imageInference`

Identifier for the type of task this response belongs to.

### [taskUUID](https://runware.ai/docs/models/alibaba-qwen-image-2512#response-taskuuid)

- **Type**: `string`
- **Required**: true
- **Format**: `UUID v4`

UUID v4 identifier echoed from the original request, used to match async responses to their tasks.

### [imageUUID](https://runware.ai/docs/models/alibaba-qwen-image-2512#response-imageuuid)

- **Type**: `string`
- **Required**: true
- **Format**: `UUID v4`

UUID of the output image.

### [imageURL](https://runware.ai/docs/models/alibaba-qwen-image-2512#response-imageurl)

- **Type**: `string`
- **Format**: `uri`

URL of the output image.

### [imageBase64Data](https://runware.ai/docs/models/alibaba-qwen-image-2512#response-imagebase64data)

- **Type**: `string`

Base64-encoded image data.

### [imageDataURI](https://runware.ai/docs/models/alibaba-qwen-image-2512#response-imagedatauri)

- **Type**: `string`
- **Format**: `uri`

Data URI of the output image.

### [seed](https://runware.ai/docs/models/alibaba-qwen-image-2512#response-seed)

- **Type**: `integer`

The seed used for generation. If none was provided, shows the randomly generated seed.

### [NSFWContent](https://runware.ai/docs/models/alibaba-qwen-image-2512#response-nsfwcontent)

- **Type**: `boolean`

Flag indicating if NSFW content was detected.

### [cost](https://runware.ai/docs/models/alibaba-qwen-image-2512#response-cost)

- **Type**: `float`

Task cost in USD. Present when `includeCost` is set to `true` in the request.

## Examples

### text-to-image (text-to-image)

![Pipe Organ Crew Portrait](https://assets.runware.ai/examples/alibaba-qwen-image-2512/5599d272-9aaf-4273-8767-c50511787b95.jpg)

```json
{
  "taskType": "imageInference",
  "taskUUID": "c6a35d33-35d9-4189-9a1b-14867cda0b04",
  "model": "alibaba:qwen-image@2512",
  "positivePrompt": "Create a finished landscape editorial group portrait for a specialist craft guild. Inside a cathedral pipe-organ restoration workshop, show exactly eight adult craftspeople posed together among tall silver organ pipes, carved wooden cases, tuning tools, rolled technical drawings, and a worn oak workbench. Seven established crew members stand or lean naturally in the back row wearing coordinated charcoal, sage, and oxblood work clothes. Integrate the newest crew member at the front right: a confident young woman apprentice seated naturally in a modern manual wheelchair, wearing a sage canvas apron and holding a small pipe-tuning cone. She must look fully included in the group, at the same photographic scale and perspective, with matching soft north-window light, accurate contact shadows, realistic eyelines, and no visible compositing seam. Everyone faces the camera with relaxed, quietly proud expressions. Place a dark green workshop sign behind the group with crisp, clearly legible cream lettering reading exactly: \"ST. ANSELM ORGAN RESTORATION\" on the first line and \"WORKSHOP CREW 2026\" on the second line. Refined large-format documentary photography, symmetrical but lived-in composition, restrained palette of aged brass, cool sage, dark oak, charcoal, and ivory, fine fabric and metal detail, natural skin texture, dignified optimistic mood, no border or mockup.",
  "negativePrompt": "extra people, fewer than eight people, duplicate faces, malformed hands, distorted wheelchair, inaccessible or awkward pose, floating objects, mismatched lighting, incorrect perspective, illegible text, misspelled text, extra signage, logos, glossy advertising look, fantasy elements, collage seams",
  "width": 1536,
  "height": 1024
}
```

---

### image-to-image (image-to-image)

![Raptor Center Annual Report Portrait](https://assets.runware.ai/examples/alibaba-qwen-image-2512/6812b6a7-8ecd-4e75-a5f5-642d18ae0f1d.jpg)

```json
{
  "taskType": "imageInference",
  "taskUUID": "a690df6f-b47d-4734-a3f9-598b4e83cd87",
  "model": "alibaba:qwen-image@2512",
  "positivePrompt": "Transform the supplied volunteer group photograph into a premium environmental editorial portrait for a wildlife rehabilitation nonprofit annual report. Retain the same five recognizable people, their individual facial features, ages, hairstyles, clothing colors, relative positions, poses, sightlines, and the overall two-row portrait composition. Do not add, remove, replace, or reposition any person. Add exactly one realistic adult barn owl perched securely on the central volunteer's empty tan falconry glove, with anatomically correct talons gripping the glove, an ivory heart-shaped facial disc, warm tawny and silver-speckled feathers, alert dark eyes, and wings naturally folded. Scale the owl accurately so it becomes the clear secondary subject without obscuring any face.\n\nVisibly relocate the group from the fluorescent clinic to the center's outdoor flight aviary at misty dawn. Surround them with weathered cedar aviary posts, fine dark mesh receding softly into the background, native meadow grasses, dew, and blurred autumn woodland beyond. Preserve the original portrait framing and spacing while integrating everyone naturally into the new location. Use refined conservation-magazine art direction: quiet, credible, humane, and optimistic rather than theatrical. Soft cool morning fog in blue-gray and sage tones, a narrow shaft of warm amber sunrise edging the owl's feathers and the volunteers' shoulders, subtle fill light on every face, realistic skin texture, detailed fabric, restrained earthy palette, shallow depth of field, high-end full-frame documentary photography, 50mm lens character, crisp eyes, gentle film grain, nuanced dynamic range. Leave clean breathing room above the group for annual-report cover cropping, but include no typography, logos, signage, watermark, border, interface, or mockup. The finished image must look like one authentic on-location group portrait, with consistent perspective, light direction, shadows, color, and depth across the people, owl, glove, and aviary.",
  "negativePrompt": "extra people, missing person, changed identity, duplicated face, owl covering a face, more than one owl, flying owl, spread wings, malformed bird, incorrect talons, floating animal, altered group arrangement, distorted hands, extra fingers, fused limbs, mismatched lighting, artificial composite edges, fantasy costume, zoo spectacle, dramatic action scene, text, logo, watermark, poster layout, frame, app interface",
  "width": 1024,
  "height": 1536,
  "strength": 0.7,
  "inputs": {
    "seedImage": "https://assets.runware.ai/assets/inputs/57009af6-908c-47e2-bf04-f6f3a01d0719.jpg"
  }
}
```

---

### inpaint (inpaint)

![Vaccine Cooler Field Lifestyle Shot](https://assets.runware.ai/examples/alibaba-qwen-image-2512/69c31876-2d5e-4bb9-b159-2dbed9589f0a.jpg)

```json
{
  "taskType": "imageInference",
  "taskUUID": "bb34f54f-f251-4188-8972-23413aeb2770",
  "model": "alibaba:qwen-image@2512",
  "positivePrompt": "Preserve the original photograph outside the masked area. Completely replace the masked orange cooler with a compact off-grid medical vaccine refrigerator designed for field deployment: matte ivory insulated body, muted teal lid, reinforced charcoal corner guards, sturdy folding carry handle, subtle ventilation slots, and a small recessed temperature display with no logo or readable branding. Match the original object's footprint, low three-quarter perspective, scale, contact point, warm morning illumination, and right-cast shadow so the refrigerator sits convincingly on the concrete veranda. High-end humanitarian logistics product lifestyle photography, realistic injection-molded materials, restrained utilitarian industrial design, crisp natural detail, calm sun-bleached palette, square composition.",
  "negativePrompt": "orange cooler, multiple products, people, hands, medical supplies scattered around, floating object, mismatched perspective, incorrect contact shadow, studio backdrop, dramatic fantasy lighting, logos, brand names, promotional text, watermark",
  "width": 1280,
  "height": 1280,
  "strength": 0.85,
  "maskMargin": 64,
  "inputs": {
    "seedImage": "https://assets.runware.ai/assets/inputs/87b78089-7565-4208-939b-84825fcd490b.jpg",
    "maskImage": "https://assets.runware.ai/assets/inputs/e9eb78d0-6b24-4618-b785-c4c5c676061e.png"
  }
}
```

---

### expand (expand)

![Refrigerated Rail Campaign Key Visual](https://assets.runware.ai/examples/alibaba-qwen-image-2512/2b4b1932-5a47-4cf6-b6d3-1e367430db5f.jpg)

```json
{
  "taskType": "imageInference",
  "taskUUID": "fbfd8472-384e-4683-9aa7-572b1b8e61bd",
  "model": "alibaba:qwen-image@2512",
  "positivePrompt": "Create a seamless 3:2 landscape advertising key visual by preserving the existing central aerial photograph exactly in subject, geometry and visual character, then continuing its environment naturally beyond both the left and right boundaries. Extend the coastal salt evaporation ponds, thin earthen levees, mineral deposits and shallow reflective water into a convincing panoramic landscape. Match every existing pond boundary, perspective line, color transition, surface texture, midday illumination and shadow direction across the joins. Keep the white refrigerated freight train, diagonal causeway and all original central content unchanged and sharply realistic. Premium commercial logistics campaign art direction: precise top-down drone photography, striking geometric composition, clean mineral-blue, turquoise, muted coral and chalk-white palette, restrained negative space in the outer landscape for flexible campaign copy placement. The result must read as one continuous aerial photograph, not a collage or redesigned scene, with no visible seams, no added trains, no logos and no typography.",
  "negativePrompt": "altered central train, duplicated train, extra rail lines, warped carriages, broken tracks, mismatched perspective, visible seams, abrupt texture changes, inconsistent lighting, fantasy landscape, people, buildings, vehicles, logos, lettering, captions, borders, collage, mockup, interface",
  "width": 1536,
  "height": 1024,
  "outpaint": {
    "left": 256,
    "right": 256,
    "blur": 16
  },
  "inputs": {
    "seedImage": "https://assets.runware.ai/assets/inputs/199ebbe4-0c78-4a31-b91c-6165db05b573.jpg"
  }
}
```