Call Go operations to generate AI portraits - Platform For AI

You can use AI Portrait SDK for Go provided by Platform for AI (PAI) to call the algorithm service operations of the AI Portrait service to train models and create portraits. You can also use the SDK to configure a custom Low-Rank Adaptation (LoRA) model and create portraits based on a template image. This topic describes the required preparations before you can use the AI Portrait SDK for Go to call the related operations. This topic also provides examples on how to call the operations.

Prerequisites

A Go environment is prepared.
For model training and portrait creation, 5 to 20 training images and 1 template image are prepared. The following image formats are supported: .jpg, .jpeg, and .png. The image size must be greater than 512 x 512 pixels.
- For a solo portrait, you only need a template image that contains the face of a person. The faces in multiple training images belong to the same person.
- For a multi-person portrait, the template image must contain multiple faces, and the number of the faces must be the same as the value of the model_id parameter used for model training.

Preparations

Environment dependency: Go 1.18 or later.
Install SDK for Go:
- Run the go mod init project name command to create a project.
- Run the go get github.com/aliyun/aliyun-pai-aiservice-go-sdk command to obtain the code of the SDK.
- Run the import "github.com/aliyun/aliyun-pai-aiservice-go-sdk" command to introduce the SDK in the code.

Initialize the client:

Run the following shell command to write the HOST, AppId, and Token parameters to environment variables.

export HOST="http://ai-service.ce8cc13b6421545749e7b4605f3d02607.cn-hangzhou.alicontainer.com"
export AppId=xxx
export Token=xxxxxxxxx

Parameter	Description
HOST	The server address. Example: `http://ai-service.ce8cc13b6421545749e7b4605f3d02607.cn-hangzhou.alicontainer.com`.
AppId	The application ID. After you activate the AI Portrait service, you can view the application ID on the AI Portrait page.
Token	The token. After you activate the AI Portrait service, you can view the token on the AI Portrait page.

Run the following code to initialize the client.

host := os.Getenv("Host")
appId := os.Getenv("AppId")
token := os.Getenv("Token")

client := sdk.NewClient(host, appId, token)

Sample code

AI Portrait is a resource-intensive service, involving the model training and portrait creation processes. In most cases, model training requires several minutes, whereas portrait creation requires only dozens of seconds. The following figure shows the workflow of API calls.

The following section describes the sample code for operation requests, responses, and end-to-end requests.

Check request (client.AIGCApi.AigcImageCheck)

Sample request

package main

import (
	"fmt"
	"log"
	"os"

	sdk "github.com/aliyun/aliyun-pai-aiservice-go-sdk"
)

func main() {
  // Obtain the HOST, AppId, and Token parameters from environment variables.
	host := os.Getenv("Host")
	appId := os.Getenv("AppId")
	token := os.Getenv("Token")

	client := sdk.NewClient(host, appId, token)

	// Check images.
	images := []string{"https://train/0.jpg","https://train/1.jpg","https://train/2.jpg","https://train/3.jpg"}

	response, err := client.AIGCApi.AigcImageCheck(images, "", nil)

	fmt.Println(response)

	if err != nil {
		log.Fatal(err)
		return
	}

	if response.Code != "OK" {
		log.Fatal("response error:%v", response)
		return
	}

	// Traverse the image check results.
	for _, result := range response.Data.CheckResults {
		fmt.Printf("code=%d\tfrontal=%v\turl=%s\n", result.Code, result.Frontal, result.Url)
	}
}

The following table describes the parameters in the preceding code.

Parameter	Description	Type
images	The URLs of the images that you want to check. Separate multiple URLs with commas (,).	[]string

Sample response

An AIGCImageCheckResponse object is returned. The following table describes the fields of the object.

Field	Description	Type
RequestId	The request ID.	string
Code	The status code of the request. Valid values: OK and error. OK indicates that the request is completed and error indicates that the request is not completed.	string
Message	The details of the request status. If the request is successful, success is returned. In other scenarios, a different message is returned.	string
Data	The details of the returned data.	AIGCImageCheckData

The following table describes the data field whose type is AIGCImageCheckData.

Parameter	Description	Type
CheckResults	The detection result of each input image. Each image corresponds to a dictionary that contains the following keys: url, message, and frontal. The url key indicates the URL of the image, the message key indicates the detection details of the image, and the frontal key indicates whether the image is front-facing.	List<AIGCImageCheckResult>
CostTime	The time required for the server to call the API.	float64
Images	The URLs of the images that are checked.	List<String>
RequestId	The ID of the request. The ID is the same as the value of the requestId parameter in the sample response.	String

The following table describes the check_results field whose type is AIGCImageCheckResult.

Parameter	Description	Type
Code	The status code of the request. Valid values: 0 and 1. 0 indicates that the request is not completed and 1 indicates that the request is completed. The detection result of each input image.	int
Frontal	Indicates whether the image is the front-facing.	bool
Message	The details of the request status.	string
Url	The URL of the image.	string

The following table describes the messages of the check_results field.

Message	Status code	Description
success	1	The image meets the requirements.
Image decode error.	2	The image cannot be downloaded or decoded.
Number of face is not 1.	3	The number of faces is not 1.
Image detect error.	4	An error occurred while detecting the face.
Image encoding error.	5	An error occurred while encoding a face as a feature vector. In most cases, the issue occurs because a face cannot be detected.
This photo is not the same person in photos.	6	If only this error is reported, the faces in multiple images do not belong to the same person.

Model training initiation (client.AIGCApi.AigcImagesTrain)

Stable Diffusion 1.5 training

package main

import (
	"fmt"
	"log"
	"os"
	"time"

	sdk "github.com/aliyun/aliyun-pai-aiservice-go-sdk"
)

func main() {
  // Obtain the HOST, AppId, and Token parameters from environment variables.
	host := os.Getenv("Host")
	appId := os.Getenv("AppId")
	token := os.Getenv("Token")

	client := sdk.NewClient(host, appId, token)

	images := []string{"https://train/0.jpg","https://train/1.jpg","https://train/2.jpg","https://train/3.jpg"}

	response, err := client.AIGCApi.AigcImagesTrain(images, "", nil)
	if err != nil {
		log.Fatal(err)
	}

	// The task ID.
	jobId := response.Data.JobId
	modelId := response.Data.ModelId
	fmt.Println(jobId)

	jobResponse, err := client.JobApi.GetAsyncJobWithId(jobId)
	fmt.Println(jobResponse.Data.Job.Message)
	if err != nil {
		log.Fatal(err)
	}

	for {
		jobResponse, err := client.JobApi.GetAsyncJobWithId(jobId)
		fmt.Println(jobResponse.Data.Job.Result)
		if err != nil {
			log.Fatal("get job fail", err)
		}

		if jobResponse.Data.Job.State == sdk.JOB_STATE_WAIT {
			fmt.Println("job running")
		} else if jobResponse.Data.Job.State == sdk.JOB_STATE_SUCCESS { // job success
			fmt.Println(jobResponse)
			fmt.Println("job success")
			break
		} else if jobResponse.Data.Job.State == sdk.JOB_STATE_FAILED {
			log.Fatal("job fail", err)
			return
		}

		time.Sleep(30 * time.Second)
	}
	fmt.Println(modelId)

}

Parameter	Description	Type
images	The URLs of the training images.	[]string

Stable Diffusion XL (SDXL) training

If you want to use SDXL, contact your account manager to activate the service and then use the service by specifying a model in the modelName parameter.

You can use SDXL for portrait training or scene LoRA training by specifying the parameters in configure.

package main

import (
	"fmt"
	"log"
	"os"
	"time"

	sdk "github.com/aliyun/aliyun-pai-aiservice-go-sdk"
)

func main() {
  // Obtain the HOST, AppId, and Token parameters from environment variables.
	host := os.Getenv("Host")
	appId := os.Getenv("AppId")
	token := os.Getenv("Token")

	client := sdk.NewClient(host, appId, token)

	images := []string{"https://train/0.jpg","https://train/1.jpg","https://train/2.jpg","https://train/3.jpg"}

	modelName := "train_xl"
	configure := make(map[string]interface{}, 10)
	configure["train_scene_lora_bool"] = false
	configure["scene_lora_prompts"] = []string{"a photography of a woman with long blonde hair and a white dress",
		"a photography of a woman in a pink dress posing for a picture",
		"a photography of a woman in a black dress with a white background",
		"a photography of a woman with a frilly collar and suspenders",
		"a photography of a woman with a white dress and a white headpiece"}

	response, err := client.AIGCApi.AigcImagesTrain(images, modelName, configure)
	if err != nil {
		log.Fatal(err)
	}

	// The task ID.
	jobId := response.Data.JobId
	modelId := response.Data.ModelId
	fmt.Println(jobId)

	jobResponse, err := client.JobApi.GetAsyncJobWithId(jobId)
	fmt.Println(jobResponse.Data.Job.Message)
	if err != nil {
		log.Fatal(err)
	}

	for {
		jobResponse, err := client.JobApi.GetAsyncJobWithId(jobId)
		fmt.Println(jobResponse.Data.Job.Result)
		if err != nil {
			log.Fatal("get job fail", err)
		}

		if jobResponse.Data.Job.State == sdk.JOB_STATE_WAIT {
			fmt.Println("job running")
		} else if jobResponse.Data.Job.State == sdk.JOB_STATE_SUCCESS { // job success
			fmt.Println(jobResponse)
			fmt.Println("job success")
			break
		} else if jobResponse.Data.Job.State == sdk.JOB_STATE_FAILED {
			log.Fatal("job fail", err)
			return
		}

		time.Sleep(30 * time.Second)
	}
	fmt.Println(modelId)

}

Parameter	Description	Type
images	The URLs of the training images.	[]string
modelName	The name of the model. If you want to use SDXL, set the value to train_xl.	string
configure	The parameter configuration. Default value: nil.	map[string]interface{}

Parameter in the configure field

Description

Type

Required

Default value

Valid value

train_scene_lora_bool

Specifies whether to perform LoRA training on the scene.

bool

False

True and False

scene_lora_prompts

The prompts of LoRA training on the scene.

If you perform LoRA training on the scene, this parameter is required. The length of the list must be the same as the length of the image URL.

[]string

[]

N/A

Response description

An AIGCImageTrainResponse object is returned. The following table describes the fields of the object.

Field	Description	Type
RequestId	The ID of the request.	string
Code	The status code of the request. Valid values: OK and error. OK indicates that the request is completed and error indicates that the request is not completed.	string
Message	The details of the request status. If the request is successful, success is returned. In other scenarios, a different message is returned.	string
Data	The details of the returned data.	AIGCImageTrainData

The following table describes the data field whose type is AIGCImageTrainData.

Parameter	Description	Type
jobId	The task ID.	int32
modelId	The ID of the model that is trained, which is a string of 36 characters.	string

You need to use job_id to query training results.

You need to use model_id to send a request for portrait creation.

Training result query (client.JobApi.GetAsyncJobWithId)

Sample request

package main

import (
	"fmt"
	"log"
	"os"
	"time"

	sdk "github.com/aliyun/aliyun-pai-aiservice-go-sdk"
)

func main() {
	// Obtain the HOST, AppId, and Token parameters from environment variables.
	host := os.Getenv("Host")
	appId := os.Getenv("AppId")
	token := os.Getenv("Token")

	client := sdk.NewClient(host, appId, token)

	// The task ID.
	var jobId int32 = xxxxxx

	for {
		jobResponse, err := client.JobApi.GetAsyncJobWithId(jobId)
		if err != nil {
			log.Fatal("get job fail", err)
		}

		if jobResponse.Data.Job.State == sdk.JOB_STATE_WAIT {
			fmt.Println("job running")
		} else if jobResponse.Data.Job.State == sdk.JOB_STATE_SUCCESS { // job success
			fmt.Println(jobResponse)
			fmt.Println("job success")
			break
		} else if jobResponse.Data.Job.State == sdk.JOB_STATE_FAILED {
			log.Fatal("job fail", err)
			return
		}

		time.Sleep(30 * time.Second)
	}

}

The following table describes the parameters in the preceding code.

Parameter	Description	Type	Required
jobId	The task ID returned in the request response.	int32	Yes

Response description

An AsyncJobResponse object is returned. The following table describes the fields in the object.

Field	Description	Type
RequestId	The ID of the request.	string
Code	The status code of the request. Valid values: OK and error. OK indicates that the request is completed and error indicates that the request is not completed.	string
Message	The details of the request status. If the request is successful, success is returned. In other scenarios, a different message is returned.	string
Data	The details of the returned data.	AsyncJobData

The following table describes the data field whose type is AsyncJobData.

Parameter	Description	Type
State	The check result of each image.	int32
AppId	The application ID of the user.	string
Message	The details of the request status. If the request is successful, success is returned. In other scenarios, a different message is returned.	string
Result	The response of the model.	string

The following table describes the parameters in the Result field returned after the model training is completed.

Parameter	Description
cost_time	The time required for this training session.
states	The check result of each image. Each image corresponds to a dictionary, and each dictionary has three keys: url, message, and frontal. The url key indicates the URL of the image, the message key indicates the detection details of the image, and the frontal key indicates whether the image is front-facing.
model_id	The ID of the LoRA model, which is the same as the value of model_id obtained during the training request.

Portrait creation

Solo portrait creation (client.AIGCApi.AigcImagesCreate)

Stable Diffusion 1.5 prediction

package main

import (
	"encoding/base64"
	"log"
	"os"

	sdk "github.com/aliyun/aliyun-pai-aiservice-go-sdk"
)

func main() {
  // Obtain the HOST, AppId, and Token parameters from environment variables.
	host := os.Getenv("Host")
	appId := os.Getenv("AppId")
	token := os.Getenv("Token")

	client := sdk.NewClient(host, appId, token)

	modelID := "xxxxxxxxxxxxxxx"

	template_image := "https://template_case.png"

	config := make(map[string]interface{}, 10)
	config["lora_weights"] = 0.9
	config["first_denoising_strength"] = 1
	config["second_denoising_strength"] = 1
	config["more_like_me"] = 1
	config["crop_face_preprocess"] = false
	config["apply_face_fusion_before"] = false
	config["apply_face_fusion_after"] = false
	config["color_shift_middle"] = false
	config["color_shift_last"] = false
	config["background_restore"] = true

	response, err := client.AIGCApi.AigcImagesCreate(modelID, image, "", config)

	if err != nil {
		log.Fatal(err)
	}

	image_decode, _ := base64.StdEncoding.DecodeString(response.Data.Image)

	f, err_2 := os.Create("test.jpg")
	f.Write(image_decode)

	if err_2 != nil {
		log.Fatal(err)
	}

}

Parameter	Description	Type	Required
modelID	The ID of the LoRA model. Enter the value of model_id obtained from training.	string	Yes
template_image	The URL path of the template.	string	Yes
config	The returned configurations. Default value: nil.	map[string]interface{}	Yes

The following table describes the parameters in the config field.

Parameter	Description	Type	Required	Default value	Valid value
lora_weights	The weight of LoRA.	float	No	0.90	0.5 to 1.0
first_denoising_strength	The intensity of the first image reconstruction.	float	No	0.45	0.0 to 1.0
second_denoising_strength	The intensity of the second image reconstruction.	float	No	0.30	0.0 to 1.0
more_like_me	The degree to which the image resembles the desired person.	float	No	0.50	0.0 to 1.0
crop_face_preprocess	Specifies whether to crop and then reconstruct the image. We recommend that you set this parameter to True for large images.	bool	No	True	True and False
apply_face_fusion_before	Specifies whether to perform the first portrait merging.	bool	No	True	True and False
apply_face_fusion_after	Specifies whether to perform the second portrait merging.	bool	No	True	True and False
color_shift_middle	Specifies whether to perform the first color correction.	bool	No	True	True and False
color_shift_last	Specifies whether to perform the second color correction.	bool	No	True	True and False
background_restore	Specifies whether to rebuild the background.	bool	No	False	True and False
skin_retouching_bool	Specifies whether to perform skin smoothing. If you set the parameter to True, the system smooths and brightens the skin. In most cases, this makes the image more attractive, but may excessively lighten the skin. If you set the parameter to False, the skin texture can be improved.	bool	No	False	True and False
photo_enhancement_bool	Specifies whether to enhance the portrait. If you set the parameter to True, the system restores the portrait or increases the image resolution to improve image quality.	bool	No	True	True and False
photo_enhancement_method	The method used to enhance the portrait. photo_fix is used for image restoration by fixing specific imperfections, which may result in a loss of skin texture due to some distortion. super_resolution is used for enhancing the resolution of an image, preserving more of the original appearance.	string	No	photo_fix	photo_fix and super_resolution
makeup_transfer	Specifies whether to perform makeup transfer. If you set the parameter to True, the system performs makeup transfer to prevent the image from being too plain. However, this may result in less resemblance of the image to the user.	bool	No	False	True and False
makeup_transfer_ratio	The intensity of makeup transfer. The larger the value, the higher the proportion of makeup transfer and the more closely the resulting makeup resembles the template. This may also result in less resemblance of the image to the user.	float	No	0.5	0.0 to 1.0
face_shape_match	Specifies whether to perform face shape compatibility. If you set the parameter to True, the face shape and skin texture more closely resemble the user.	bool	No	False	True and False
ipa_control	Specifies whether to use IP-Adapter. If you set the parameter to True, the generated portrait more closely resembles the reference image.	bool	No	False	True and False
ipa_control_only	If you set the parameter to True, you must configure the ipa_image_path parameter and you can directly perform prediction without training.	bool	No	False	True and False
ipa_image_path	The URL of the reference image. If you set the ipa_control_only parameter to True, you must configure this parameter.	string	No	None	A publicly accessible URL
ipa_weight	The weight of IP-Adapter. The larger the value is, the more closely the generated image resembles the user. However, an excessively large value may result in image distortion.	float	No	0.5	0.0 to 1.0
style_name	The style of the created portrait.	string	No	Realistic	Realistic and Anime
lcm_accelerate	Specifies whether to perform latent code manipulation (LCM) acceleration.	bool	No	False	True and False
sharp_ratio	The sharpness level. You can configure this parameter to enhance clarity. An excessively large value may result in image distortion.	float	No	0.15	0.0 to 1.0
t2i_prompt	The prompt. Configure this parameter if you want to use the text-to-image generation feature. If you configure the t2i_prompt parameter, you do not need to specify the URL path of the template.	String	No	None	N/A

SDXL prediction

If you want to use SDXL, contact your account manager to activate the service and then use the service by specifying a model in the modelName parameter.

package main

import (
	"encoding/base64"
	"log"
	"os"

	sdk "github.com/aliyun/aliyun-pai-aiservice-go-sdk"
)

func main() {
  // Obtain the HOST, AppId, and Token parameters from environment variables.
	host := os.Getenv("Host")
	appId := os.Getenv("AppId")
	token := os.Getenv("Token")

	client := sdk.NewClient(host, appId, token)

	modelID := "xxxxxxxxxxxx"

	template_image := "https://template_case.png"

	modelName := "create_xl"
	config := make(map[string]interface{}, 10)
	// config["lora_weights"] = 0.9
	// config["first_denoising_strength"] = 1
	// config["second_denoising_strength"] = 1
	// config["more_like_me"] = 1
	// config["crop_face_preprocess"] = false
	// config["apply_face_fusion_before"] = false
	// config["apply_face_fusion_after"] = false
	// config["color_shift_middle"] = false
	// config["color_shift_last"] = false
	// config["background_restore"] = true

	response, err := client.AIGCApi.AigcImagesCreate(modelID, image, modelName, config)

	if err != nil {
		log.Fatal(err)
	}

	image_decode, _ := base64.StdEncoding.DecodeString(response.Data.Image)

	f, err_2 := os.Create("test.jpg")
	f.Write(image_decode)

	if err_2 != nil {
		log.Fatal(err)
	}

}

Parameter	Type	Description
modelID	string	The ID of the LoRA model. Enter the value of model_id obtained from training. Set the value to "" if you set the ipa_control_only parameter to True.
template_image	string	The URL path of the template image. If you use scene_lora or prompt to generate an image, set the parameter to t2i_generate.
modelName	string	The name of the model. If SDXL is used, set the value to create_xl.
config	map[string]interface{}	The returned configuration of the model. Default value: None.

The following table describes the parameters in the config field.

Parameter	Description	Type	Required	Default value	Valid value
lora_weights	The weight of LoRA. The larger the value is, the more closely the generated image resembles the user. However, an excessively large value may result in image distortion.	float	No	0.90	0.5 to 1.0
first_diffusion_steps	The number of steps for the first diffusion. We recommend that you keep the parameter value unchanged. A significant decrease can cause the image to become distorted.	int	No	50	20 to 50
first_denoising_strength	The intensity of the first image reconstruction. The reconstruction intensity of the face. A greater value specifies the greater intensity of the face reconstruction, and the more closely the generated image resembles the user. However, a large value may result in an uncoordinated image.	float	No	0.45	0.0 to 1.0
second_diffusion_steps	The number of steps for the second diffusion. We recommend that you keep the parameter value unchanged. A significant decrease can cause the image to become distorted.	int	No	30	20 to 50
second_denoising_strength	The intensity of the second image reconstruction. The reconstruction intensity of the facial edges. An excessive large value may result in an uncoordinated image.	float	No	0.30	0.0 to 1.0
more_like_me	The degree to which the image resembles the desired person. The larger the value is, the more closely the generated image resembles the desired person. However, if the value is excessively large, the image may look unrealistic.	float	No	0.60	0.0 to 1.0
crop_face_preprocess	Specifies whether to crop and then reconstruct the image. We recommend that you set this parameter to True for large images. We recommend that you keep the parameter value unchanged.	bool	No	True	True and False
apply_face_fusion_before	Specifies whether to perform the first portrait merging. If you set the parameter to True, the system performs portrait merging. If you set the parameter to False, the similarity is reduced.	bool	No	True	True and False
apply_face_fusion_after	Specifies whether to perform the second portrait merging. If you set the parameter to True, the system performs portrait merging. If you set the parameter to False, the similarity is reduced.	bool	No	True	True and False
color_shift_middle	Specifies whether to perform the first color correction. If you set the parameter to True, the system performs color correction to increase the resemblance of the skin color of the generated image to the template. If you set the parameter to False, the system does not correct colors and color distortion may occur.	bool	No	True	True and False
color_shift_last	Specifies whether to perform the second color correction. If you set the parameter to True, the system performs color correction to increase the resemblance of the skin color of the generated image to the template. If you set the parameter to False, the system does not correct colors and color distortion may occur.	bool	No	True	True and False
background_restore	Specifies whether to rebuild the background. If you set the parameter to True, the background is reconstructed to be more natural. However, this feature changes the image background and increases the time consumption.	bool	No	False	True and False
skin_retouching_bool	Specifies whether to perform skin smoothing. If you set the parameter to True, the system smooths and brightens the skin. In most cases, this makes the image more attractive, but may excessively lighten the skin. If you set the parameter to False, the skin texture can be improved.	bool	No	False	True and False
photo_enhancement_bool	Specifies whether to enhance the portrait. If you set the parameter to True, the system restores the portrait or increases the image resolution to improve image quality.	bool	No	True	True and False
photo_enhancement_method	The method to enhance the portrait. photo_fix is used for image restoration by fixing specific imperfections, which may result in a loss of skin texture due to some distortion. super_resolution is used for enhancing the resolution of an image, preserving more of the original appearance.	String	No	photo_fix	photo_fix and super_resolution
makeup_transfer	Specifies whether to perform makeup transfer. If you set the parameter to True, the system performs makeup transfer to prevent the image from being too plain. However, this may result in less resemblance of the image to the user.	bool	No	False	True and False
makeup_transfer_ratio	The intensity of makeup transfer. The larger the value, the higher the proportion of makeup transfer and the more closely the resulting makeup resembles the template. This may also result in less resemblance of the image to the user.	float	No	0.5	0.0 to 1.0
ipa_control	Specifies whether to use IP-Adapter. If you set the parameter to True, the generated portrait more closely resembles the reference image.	bool	No	False	True and False
ipa_control_only	If you set the parameter to True, you must configure the ipa_image_path parameter and you can directly perform prediction without training.	bool	No	False	True and False
ipa_image_path	The URL of the reference portrait. If you set the ipa_control_only to True, you must configure this parameter.	String	No	None	A publicly accessible URL
ipa_weight	The weight of IP-Adapter. The larger the value is, the more closely the generated image resembles the user. However, an excessively large value may result in image distortion.	float	No	0.5	0.0 to 1.0
lcm_accelerate	Specifies whether to perform LCM acceleration.	bool	No	False	True and False
sharp_ratio	The sharpness level. You can configure this parameter to enhance clarity. An excessively large value may result in image distortion.	float	No	0.15	0.0 to 1.0
scene_id	If you want to use scene_lora, you need to configure the scene_id parameter. The scene ID is obtained by training. The model ID that is returned when you train the SDXL_LoRA model is the scene ID.	String	No	None	N/A
t2i_prompt	The prompt. Configure this parameter if you want to use the text-to-image generation feature. If you configure the t2i_prompt parameter, you do not need to specify the URL path of the template.	String	No	None	N/A

Multi-person portrait creation (client.AIGCApi.AigcImagesCreateByMultiModelIds)

package main

import (
	"encoding/base64"
	"log"
	"os"

	sdk "github.com/aliyun/aliyun-pai-aiservice-go-sdk"
)

func main() {
  // Obtain the HOST, AppId, and Token parameters from environment variables.
	host := os.Getenv("Host")
	appId := os.Getenv("AppId")
	token := os.Getenv("Token")

	client := sdk.NewClient(host, appId, token)

	modelIDs := []string{"xxxxxx", "xxxxxx"}

	template_image := "https://template_case.png"

	config := make(map[string]interface{}, 10)
	config["lora_weights"] = 0.9
	config["first_denoising_strength"] = 1
	config["second_denoising_strength"] = 1
	config["more_like_me"] = 1
	config["crop_face_preprocess"] = false
	config["apply_face_fusion_before"] = false
	config["apply_face_fusion_after"] = false
	config["color_shift_middle"] = false
	config["color_shift_last"] = false
	config["background_restore"] = true

	response, err := client.AIGCApi.AigcImagesCreateByMultiModelIds(modelIDs, image, "", config)

	image_decode, _ := base64.StdEncoding.DecodeString(response.Data.Image)

	f, err := os.Create("test.jpg")
	f.Write(image_decode)

	if err != nil {
		log.Fatal(err)
	}

}

Parameter	Description	Type	Required
modelIds	The IDs of the LoRA models that are used to create the portrait.	[]string	Yes
template_image	The URL path of the template.	string	Yes
config	The returned configurations. Default value: nil.	map[string]interface{}	Yes

Response description

An AIGCImageCreateResponse object is returned. The following table describes the fields in the object.

Field	Description	Type
RequestId	The ID of the request.	string
Code	The status code of the request. Valid values: OK and error. OK indicates that the request is completed and error indicates that the request is not completed.	string
Message	The details of the request status. If the request is successful, success is returned. In other scenarios, a different message is returned.	string
Data	The details of the returned data.	AIGCImageCreateData

The following table describes the data fields whose type is AIGCImageCreateData.

Parameter	Description	Type
CostTime	The time required for portrait creation.	Float
Image	The Base64-encoded image.	String

Description of related error codes

Error codes for service request:

HTTP status code	code	message	Description
400	PARAMETER_ERROR	not found appid	The application ID is invalid.
400	EXCEEDED_QUOTA_ERROR	exceeded quota	The quota for service calls of the account is used up.
401	PARAMETER_ERROR	sign error	The token is invalid.
404	PARAMETER_ERROR	model not found	The requested model service is not deployed.

Error codes for result queries:

HTTP status code	code	message	Description
462	error	Invalid input data. Please check the input dict.	An error occurred while parsing the input data.
462	error	Image not provided. Please check the template_image.	The template image that is used for portrait creation is not provided.
462	error	Prompts get error. Please check the model_id.	The format of the model ID is invalid.
462	error	Roop image decode error. Please check the user's lora is trained or not.	The Roop image does not exist. Check whether the model is trained.
462	error	Template image decode error. Please Give a new template.	An error occurred while decoding the template image. Provide a new template image.
462	error	There is not face in template. Please Give a new template.	No face exists in the template image. Provide a new template image.
462	error	Template image process error. Please Give a new template.	An error occurred while preprocessing the template image. Provide a new template image.
469	error	First Face Fusion Error, Can't get face in template image.	An error occurred during the first portrait merging.
469	error	First Stable Diffusion Process error. Check the webui status.	An error occurred during the first image generation by using Stable Diffusion.
469	error	Second Face Fusion Error, Can't get face in template image.	An error occurred during the second portrait merging.
469	error	Second Stable Diffusion Process error. Check the webui status.	An error occurred during the second image generation by using Stable Diffusion.
469	error	Please confirm if the number of faces in the template corresponds to the user ID.	The number of user IDs that you provided does not match the number of faces.
469	error	Third Stable Diffusion Process error. Check the webui status.	An error occurred while preprocessing the background. Provide a new template image.
500	error	Face id image decode error. Please check the user's lora is trained or not.	An error occurred while decoding the training image that you uploaded. Check whether the model is trained.

Sample code of end-to-end portrait creation

Stable Diffusion 1.5

package main

import (
	"encoding/base64"
	"fmt"
	"log"
	"os"
	"time"

	sdk "github.com/aliyun/aliyun-pai-aiservice-go-sdk"
)

func main() {
	host := os.Getenv("Host")
	appId := os.Getenv("AppId")
	token := os.Getenv("Token")

	client := sdk.NewClient(host, appId, token)
	images := []string{"https://train/0.jpg","https://train/1.jpg","https://train/2.jpg","https://train/3.jpg"}
		
	trainresponse, err := client.AIGCApi.AigcImagesTrain(images, "", nil)
	if err != nil {
		log.Fatal(err)
	}

	jobId := trainresponse.Data.JobId
	modelId := trainresponse.Data.ModelId
	fmt.Println(jobId)

	jobResponse, err := client.JobApi.GetAsyncJobWithId(jobId)
	fmt.Println(jobResponse.Data.Job.Message)
	if err != nil {
		log.Fatal(err)
	}

	for {
		jobResponse, err := client.JobApi.GetAsyncJobWithId(jobId)
		fmt.Println(jobResponse.Data.Job.Result)
		if err != nil {
			log.Fatal("get job fail", err)
		}

		if jobResponse.Data.Job.State == sdk.JOB_STATE_WAIT {
			fmt.Println("job running")
		} else if jobResponse.Data.Job.State == sdk.JOB_STATE_SUCCESS { // job success
			fmt.Println(jobResponse)
			fmt.Println("job success")
			break
		} else if jobResponse.Data.Job.State == sdk.JOB_STATE_FAILED {
			log.Fatal("job fail", err)
			return
		}

		time.Sleep(30 * time.Second)

	}
	fmt.Println(modelId)

	template_image := "https://template_case.png"

	config := make(map[string]interface{}, 10)
	config["lora_weights"] = 0.9
	config["first_denoising_strength"] = 1
	config["second_denoising_strength"] = 1
	config["more_like_me"] = 1
	config["crop_face_preprocess"] = false
	config["apply_face_fusion_before"] = false
	config["apply_face_fusion_after"] = false
	config["color_shift_middle"] = false
	config["color_shift_last"] = false
	config["background_restore"] = true

	createresponse, err := client.AIGCApi.AigcImagesCreate(modelId, template_image, "", config)
	if err != nil {
		log.Fatal(err)
	}

	image_decode, _ := base64.StdEncoding.DecodeString(createresponse.Data.Image)

	f, err := os.Create("test.jpg")
	f.Write(image_decode)

	if err != nil {
		log.Fatal(err)
	}
}

Parameter	Description	Type	Required
images	The URLs of the training images.	[]string	Yes
template_image	The URL path of the template.	string	Yes
config	The returned configurations. Default value: nil.	map[string]interface{}	Yes

SDXL

package main

import (
	"encoding/base64"
	"fmt"
	"log"
	"os"
	"time"

	sdk "github.com/aliyun/aliyun-pai-aiservice-go-sdk"
)

func main() {
	host := os.Getenv("Host")
	appId := os.Getenv("AppId")
	token := os.Getenv("Token")

	client := sdk.NewClient(host, appId, token)
	images := []string{
            "https://xxx/0.jpg",	
            "https://xxx/1.jpg",
            "https://xxx/2.jpg"}
                     
	trainresponse, err := client.AIGCApi.AigcImagesTrain(images, "train_xl", nil)
	if err != nil {
		log.Fatal(err)
	}

	jobId := trainresponse.Data.JobId
	modelId := trainresponse.Data.ModelId
	fmt.Println(jobId)

	jobResponse, err := client.JobApi.GetAsyncJobWithId(jobId)
	fmt.Println(jobResponse.Data.Job.Message)
	if err != nil {
		log.Fatal(err)
	}

	for {
		jobResponse, err := client.JobApi.GetAsyncJobWithId(jobId)
		fmt.Println(jobResponse.Data.Job.Result)
		if err != nil {
			log.Fatal("get job fail", err)
		}

		if jobResponse.Data.Job.State == sdk.JOB_STATE_WAIT {
			fmt.Println("job running")
		} else if jobResponse.Data.Job.State == sdk.JOB_STATE_SUCCESS { // job success
			fmt.Println(jobResponse)
			fmt.Println("job success")
			break
		} else if jobResponse.Data.Job.State == sdk.JOB_STATE_FAILED {
			log.Fatal("job fail", err)
			return
		}

		time.Sleep(30 * time.Second)

	}
	fmt.Println(modelId)

	template_image = "https://demo.jpg"

	config := make(map[string]interface{}, 10)

	createresponse, err := client.AIGCApi.AigcImagesCreate(modelId, template_image, "create_xl", config)
	if err != nil {
		log.Fatal(err)
	}

	image_decode, _ := base64.StdEncoding.DecodeString(createresponse.Data.Image)

	f, err := os.Create("test.jpg")
	f.Write(image_decode)

	if err != nil {
		log.Fatal(err)
	}
}

Parameter	Description	Type	Required
images	The URLs of the training images.	[]string	Yes
template_image	The URL path of the template.	string	Yes
config	The returned configurations. Default value: nil.	map[string]interface{}	Yes

AI portrait creation based on a single reference image (without the need for model training)

package main

import (
	"encoding/base64"
	"log"
	"os"

	sdk "github.com/aliyun/aliyun-pai-aiservice-go-sdk"
)

func main() {
	host := os.Getenv("Host")
	appId := os.Getenv("AppId")
	token := os.Getenv("Token")

	client := sdk.NewClient(host, appId, token)

	template_image = "https://demo.jpg"
  ref_image = "https://reference.jpg"
  
	config := make(map[string]interface{}, 10)
	config["ipa_control_only"] = true
	config["ipa_weight"] = 0.6
	config["ipa_image_path"] = ref_image

	createresponse, err := client.AIGCApi.AigcImagesCreate("", template_image, "", config)
	if err != nil {
		log.Fatal(err)
	}

	image_decode, _ := base64.StdEncoding.DecodeString(createresponse.Data.Image)

	f, err := os.Create("test.jpg")
	f.Write(image_decode)

	if err != nil {
		log.Fatal(err)
	}
}

Parameter	Description	Type	Required
template_image	The URL path of the template.	string	Yes
ref_image	The URL path of the reference image.	string	Yes
config	The returned configurations. Default value: nil.	map[string]interface{}	Yes

Template creation link based on a prompt (without the need for a template image)

package main

import (
	"encoding/base64"
	"fmt"
	"log"
	"os"
	"time"

	sdk "github.com/aliyun/aliyun-pai-aiservice-go-sdk"
)

func main() {
	host := os.Getenv("Host")
	appId := os.Getenv("AppId")
	token := os.Getenv("Token")

	client := sdk.NewClient(host, appId, token)
	images := []string{"https://xxx/0.jpg",
      							 "https://xxx/1.jpg",
                     "https://xxx/2.jpg"}
	trainresponse, err := client.AIGCApi.AigcImagesTrain(images, "", nil)
	if err != nil {
		log.Fatal(err)
	}

	jobId := trainresponse.Data.JobId
	modelId := trainresponse.Data.ModelId
	fmt.Println(jobId)

	jobResponse, err := client.JobApi.GetAsyncJobWithId(jobId)
	fmt.Println(jobResponse.Data.Job.Message)
	if err != nil {
		log.Fatal(err)
	}

	for {
		jobResponse, err := client.JobApi.GetAsyncJobWithId(jobId)
		fmt.Println(jobResponse.Data.Job.Result)
		if err != nil {
			log.Fatal("get job fail", err)
		}

		if jobResponse.Data.Job.State == sdk.JOB_STATE_WAIT {
			fmt.Println("job running")
		} else if jobResponse.Data.Job.State == sdk.JOB_STATE_SUCCESS { // job success
			fmt.Println(jobResponse)
			fmt.Println("job success")
			break
		} else if jobResponse.Data.Job.State == sdk.JOB_STATE_FAILED {
			log.Fatal("job fail", err)
			return
		}

		time.Sleep(30 * time.Second)

	}
	fmt.Println(modelId)

	template_image = "https://demo.jpg"

	t2i_prompt := "(portrait:1.5), 1girl, bokeh, bouquet, brown_hair, cloud, flower, hairband, hydrangea, lips, long_hair, outdoors, sunlight, white_flower, white_rose, green sweater, sweater, (cloth:1.0), (best quality), (realistic, photo-realistic:1.3), film photography, minor acne, (portrait:1.1), (indirect lighting), extremely detailed CG unity 8k wallpaper, huge filesize, best quality, realistic, photo-realistic, ultra high res, raw photo, put on makeup"

	config := make(map[string]interface{}, 10)
	config["t2i_prompt"] = t2i_prompt

	createresponse, err := client.AIGCApi.AigcImagesCreate(modelId, template_image, "", config)
	if err != nil {
		log.Fatal(err)
	}

	image_decode, _ := base64.StdEncoding.DecodeString(createresponse.Data.Image)

	f, err := os.Create("test.jpg")
	f.Write(image_decode)

	if err != nil {
		log.Fatal(err)
	}
}

Parameter	Description	Type	Required
images	The URLs of the images that are used for LoRA training.	[]string	Yes
template_image	The URL path of the template.	string	Yes
t2i_prompt	The prompt.	string	Yes
config	The returned configurations. Default value: nil.	map[string]interface{}	Yes

AI portrait creation based on a prompt and a template image that is generated based on a single reference image without the need for a template image or model training

package main

import (
	"encoding/base64"
	"log"
	"os"

	sdk "github.com/aliyun/aliyun-pai-aiservice-go-sdk"
)

func main() {
	host := os.Getenv("Host")
	appId := os.Getenv("AppId")
	token := os.Getenv("Token")

	client := sdk.NewClient(host, appId, token)

	template_image = "https://demo.jpg"
  ref_image = "https://reference.jpg"
  
	t2i_prompt := "(portrait:1.5), 1girl, bokeh, bouquet, brown_hair, cloud, flower, hairband, hydrangea, lips, long_hair, outdoors, sunlight, white_flower, white_rose, green sweater, sweater, (cloth:1.0), (best quality), (realistic, photo-realistic:1.3), film photography, minor acne, (portrait:1.1), (indirect lighting), extremely detailed CG unity 8k wallpaper, huge filesize, best quality, realistic, photo-realistic, ultra high res, raw photo, put on makeup"

	config := make(map[string]interface{}, 10)
	config["ipa_control_only"] = true
	config["ipa_weight"] = 0.6
	config["ipa_image_path"] = ref_image
	config["t2i_prompt"] = t2i_prompt

	createresponse, err := client.AIGCApi.AigcImagesCreate("", template_image, "", config)
	if err != nil {
		log.Fatal(err)
	}

	image_decode, _ := base64.StdEncoding.DecodeString(createresponse.Data.Image)

	f, err := os.Create("test.jpg")
	f.Write(image_decode)

	if err != nil {
		log.Fatal(err)
	}
}

Parameter	Description	Type	Required
ref_image	The URL path of the reference image.	string	Yes
template_image	The URL path of the template.	string	Yes
t2i_prompt	The prompt.	string	Yes
config	The returned configurations. Default value: nil.	map[string]interface{}	Yes

References

For more information, see GitHub.