AI portrait generation API details

AI portrait generation

AI Portrait 2.0 supports two modes: the LoRA model training mode and the training-free mode.

1) LoRA model training mode: This mode uses a pre-trained LoRA model of a character's likeness. You can then use the portrait generation model to create high-fidelity portraits of that character. This mode supports multiple preset styles, such as ID photos, business portraits, retro, and summer sports. You can also upload custom style templates to generate personalized AI portraits.

2) Training-free mode (Recommended): You can upload a set of user photos (at least one clear, front-facing photo) and a custom style template. The portrait generation model then instantly creates an AI portrait without any training. This mode only supports generating portraits using uploaded custom style templates.

LoRA model training mode details:

• Flowchart for the LoRA model training mode:

In the LoRA model training mode, AI portrait generation uses the image generation capabilities of a diffusion model and combines them with LoRA training to merge a person's likeness with a specific style. A series of post-processing steps are then applied to create portraits that are realistic, aesthetically pleasing, and closely resemble the subject. This method produces highly personalized, high-fidelity, and high-quality portraits.

The following images show examples of this feature:

• Input image

• Generated result (Business portrait)

• Preset style template

• Custom template:

• Input image

• Custom template

• Generated result

Training-free mode details:

• Flowchart for the training-free mode:

The training-free mode uses powerful, built-in pre-trained foundation models for AI portraits. This allows the diffusion model to generate images instantly without training. A series of post-processing steps are also applied to create portraits that are realistic, aesthetically pleasing, and closely resemble the subject. This method produces highly personalized, high-quality, and diverse portraits with high generation speeds.

The following images show examples of this feature:

• Input image

• Custom template

• Generated result

Scenarios

• Professional ID photo creation: You can instantly train an ID photo style for a user to create high-quality, standardized professional ID photos for various scenarios, such as corporate recruitment and resumes. This saves time, effort, and cost.

• AI portrait studio application: You can build an intelligent portrait studio application to provide users with a convenient, efficient, and highly personalized online photography experience. This lets users take high-quality, diverse personal portraits that are comparable to professional photos, anytime and anywhere.

• Global travel portrait generation: You can create an innovative digital application for generating travel portraits. This gives users the experience of having professional travel photos taken around the world. Users can obtain beautiful personal portraits in front of global landscapes and famous attractions without needing to travel.

• Commercial portrait creation: You can generate commercial portraits in batches. This is ideal for business figures, models, influencers, or corporate executives who need high-quality portraits for social media, promotional materials, magazine covers, and other commercial activities.

Features and advantages

• High-fidelity portrait creation: The face is naturally fused into the portrait. The generated portrait has a consistent facial expression, detailed features, and skin tone. It also features high definition and accurate color reproduction, resulting in superior image quality.

• Personalized style customization: This feature supports a high degree of customization with different template styles, including various scenes and shooting styles.

• Efficient character design: Compared to traditional photography, AIGC technology eliminates the need for an actual photoshoot, saving time and labor costs. It can generate many portraits in different styles in a short time, meeting the demand from consumer clients for diverse and rapid responses.

• Face quality assessment for portrait training: The service assesses whether the uploaded face images meet the standards required for portrait generation fine-tuning. It performs a multi-dimensional quality check on the number of faces, size, angle, lighting, and definition.

• Stable and easy-to-use platform service: The service provides stable portrait generation responses under high concurrency and heavy traffic, with a 99.99% reliability guarantee. It offers simple training and inference APIs that are easy to call, integrate, and highly compatible.

Model overview

Input limits

Limits for single-person photos:

• Image format: JPEG, JPG, PNG.

• Image size: No larger than 3 MB.

• Image resolution: Greater than 256 × 256 pixels and less than 2048 × 2048 pixels. The face in the image must be at least 128 × 128 pixels.

• The URL cannot contain Chinese characters.

• The image must be a clear, front-facing photo of a single person. The face angle cannot exceed 15 degrees.

• You must provide at least 1 and up to 5 input images. The images must first be verified using the face detection API (facechain-facedetect).

Limits for custom templates:

• Image format: JPEG, JPG, PNG.

• Image size: No larger than 5 MB.

• Image resolution: Greater than 512 × 512 pixels and less than 1680 × 1260 pixels.

• The URL cannot contain Chinese characters.

• The template must be a clear, high-quality image of a single person. The face must not be occluded or blurred.

curl --location 'https://{WorkspaceId}.cn-beijing.maas.aliyuncs.com/api/v1/services/aigc/album/gen_potrait' \
--header 'X-DashScope-Async: enable' \
--header 'Authorization: Bearer <your-dashscope-api-key>' \
--header 'Content-Type: application/json' \
--data '{
"model": "facechain-generation",
"parameters": {
"style": "f_idcard_female", 
"size": "768*1024",
"n":4
},
"resources": [
{
"resource_id": "women_model",
"resource_type": "facelora"
}
]
}'

curl --location 'https://{WorkspaceId}.cn-beijing.maas.aliyuncs.com/api/v1/services/aigc/album/gen_potrait' \
--header 'X-DashScope-Async: enable' \
--header 'Authorization: Bearer <your-dashscope-api-key>' \
--header 'Content-Type: application/json' \
--data '{
"model": "facechain-generation",
"parameters": {
"style": "portrait_url_template", 
"n":1
},
"input": 
{
"template_url": "<template_pic_url>"
},
"resources": [
{
"resource_id": "women_model",
"resource_type": "facelora"
}
]
}'

curl --location 'https://{WorkspaceId}.cn-beijing.maas.aliyuncs.com/api/v1/services/aigc/album/gen_potrait' \
--header 'Authorization: Bearer <your-dashscope-api-key>' \
--header 'X-DashScope-Async: enable' \
--header 'Content-Type: application/json' \
--data '{
    "model": "facechain-generation",
    "parameters":
    {
        "style": "train_free_portrait_url_template",
        "n": 1
    },
    "input":
    {
      "template_url": "<template_pic_url>",
      "user_urls": ["<user_pic_url>"]
    }
}'

{
  "request_id": "0b6a411a-56e6-9028-84eb-6852183248d8",
  "output": {
    "task_id": "7eb1032e-09a1-4df5-96ed-14091859ad57",
    "task_status": "SUCCEEDED",
    "submit_time": "2025-08-15 10:20:22.629",
    "scheduled_time": "2025-08-15 10:20:22.655",
    "end_time": "2025-08-15 10:20:50.540",
    "results": [
      {
        "url": "https://example.com/image1.jpg"
      },
      {
        "url": "https://example.com/image2.jpg"
      },
      {
        "url": "https://example.com/image3.jpg"
      },
      {
        "url": "https://example.com/image4.jpg"
      }
    ],
    "task_metrics": {
      "TOTAL": 4,
      "SUCCEEDED": 4,
      "FAILED": 0
    }
  },
  "usage": {
    "image_count": 4
  }
}

{
    "code":"InvalidApiKey",
    "message":"Invalid API-key provided.",
    "request_id":"fb53c4ec-1c12-4fc4-a580-cdb7c3261fc1"
}

curl -X GET \
--header 'Authorization: Bearer <your-dashscope-api-key>' \
https://{WorkspaceId}.cn-beijing.maas.aliyuncs.com/api/v1/tasks/86ecf553-d340-4e21-af6e-a0c6a421c010

{
    "request_id":"e5d70b02-ebd3-98ce-9fe8-759d7d7b107d",
    "output":{
        "task_id":"86ecf553-d340-4e21-af6e-a0c6a421c010",
        "task_status":"RUNNING",
        "task_metrics":{
            "TOTAL":1,
            "SUCCEEDED":1,
            "FAILED":0
        }
    }
}

{
  "request_id": "<your request id>",
  "output": {
    "task_id": "<your task id>",
    "task_status": "SUCCEEDED",
    "submit_time": "xxx",
    "scheduled_time": "xxx",
    "end_time": "xxx",
    "results": [
      {"url": "http://oss.aliyuncs.com/xxx/abc1.jpg"},
      {"url": "http://oss.aliyuncs.com/xxx/abc2.jpg"},
      {"url": "http://oss.aliyuncs.com/xxx/abc3.jpg"},
      {"url": "http://oss.aliyuncs.com/xxx/abc4.jpg"}
    ],
    "task_metrics": {
      "TOTAL": 4,
      "SUCCEEDED": 4,
      "FAILED": 0
    }
  },
  "usage": {
    "image_count": 4
  }
}

{
  "request_id": "<your request id>",
  "output": {
    "task_id": "<your task id>",
    "task_status": "FAILED",
    "submit_time": "xxx",
    "scheduled_time": "xxx",
    "end_time": "xxx",
    "code": "InvalidImageResolution",
    "message": "The input image resolution is too large or small"
  },
  "usage": {
    "image_num": 0
  }
}