美摄美颜SDK集成指南

使用提示

本文是美摄美颜SDK标准的集成指南文档。用以指导 SDK 的使用方法，默认读者已经熟悉 DevEco Studio 的基本使用方法，以及具有一定的 HarmonyOS 编程知识基础。目前支持 HarmonyOS API 12 及其以上版本。

产品功能说明

美摄美颜SDK，其包含了拍摄美颜、编辑美颜功能，美摄公司会随时根据手机系统、手机硬件、使用场景的变化，快速调整，完善，升级SDK工具包，保证SDK包的稳定性、高效率、高兼容性、给开发者带来良好的服务体验。

试用与正式使用

美摄美颜SDK免费给任何开发者提供试用机会，只需通过在美摄网站完成开发者注册，即可进行SDK下载试用，但是在生成作品中，会带有试用水印，以保证美摄公司权益，如果想获得完整的使用体验，请联系商务团队，进行沟通洽谈。

网站：www.meishesdk.com

商务邮箱：bd@meishesdk.com

主要功能

美摄美颜特效渲染主要特点：稳定、高效率、高兼容性、给开发者带来良好的服务体验。

集成方式

美摄美颜SDK提供了两种集成方式，分别是自动集成和手动集成。

依赖包说明

@meishe/libnvstreamingsdkcore：美摄美颜SDK

自动集成 har 依赖说明：

在 entry 模块下的 oh-package.json5 文件添加：

"dependencies": { 
  "@meishe/libnvstreamingsdkcore": "x.x.x"
  //填写对应版本号，如："@meishe/libnvstreamingsdkcore": "^3.14.3" 
},

说明2：现在har为字节码，需升级ide到5.0.3.500以上，并在工程级（最外层）build-profile.json5，配置"useNormalizedOHMUrl": true

"products ": [ 
  { 
    "name ": "default ", 
    "signingConfig ": "default ", 
    "compatibleSdkVersion ": "5.0.0(12) ", 
    "runtimeOS ": "HarmonyOS ", 
    "buildOption ": { 
      "strictMode ": { 
        "useNormalizedOHMUrl ": true 
      } 
    } 
  } 
],

手动集成

集成压缩包下载链接：https://www.meishesdk.com/downloadsNvStreamingSdk_Harmony_xxx.zip

集成压缩包内容

libNvStreamingSdkCore-signed.har：核心业务包
doxygen：文档
edit：是一个鸿蒙 demo 项目代码，通过这个演示了基本用法，可以用来做参考。

har 文件集成

解压缩 NvStreamingSdk_Harmony_xxx.zip 集成压缩包。
复制 libNvStreamingSdkCore-signed.har 到你的工程的外部目录下。(这个目录可以自定义，也可以放到放到工程内部)

说明：关联libNvStreamingSdkCore-signed.har，如，你复制 har 到 lib/ohos_arm64/ 目录下，工程的目录是samples/edit，sample和 lib在同一级目录，那么在 entry 模块下的 oh-package.json5 文件添加

"dynamicDependencies": { 
  "libNvStreamingSdkCore": "file:../../lib/ohos_arm64/libNvStreamingSdkCore-signed.har", 
}

说明2：现在har为字节码，需升级ide到5.0.3.500以上，并在工程级（最外层）build-profile.json5，配置"useNormalizedOHMUrl": true

"products ": [ 
  { 
    "name ": "default ", 
    "signingConfig ": "default ", 
    "compatibleSdkVersion ": "5.0.0(12) ", 
    "runtimeOS ": "HarmonyOS ", 
    "buildOption ": { 
      "strictMode ": { 
        "useNormalizedOHMUrl ": true 
      } 
    } 
  } 
],

配置权限信息

说明：sdk的拍摄和编辑过程中会使用到一些授权，可以提前配置

权限	说明	是否必须
ohos.permission.CAMERA	使用相机录制视频	是
ohos.permission.MICROPHONE	使用相机录制视频	是
ohos.permission.ACCELEROMETER	允许应用读取加速度传感器的数据	否
ohos.permission.INTERNET	sdk 在线授权验证	是

说明2：在 entry 模块下的 module.json5 文件添加，reason的字段可以在resources/base/element/string.json下配置

"requestPermissions ":[ 
  { 
    "name ": "ohos.permission.CAMERA ", 
    "reason ": "$string:camera ", 
    "usedScene ": { 
      "abilities ": [ "EntryAbility " ], 
      "when ": "always " 
    } 
  },
  { 
    "name ": "ohos.permission.MICROPHONE ", 
    "reason ": "$string:microphone ", 
    "usedScene ": { 
      "abilities ": [ "EntryAbility " ], 
      "when ": "always " 
    } 
  },
  { 
    "name ": "ohos.permission.ACCELEROMETER ", 
    "reason ": "$string:accelerometer ", 
    "usedScene ": { 
      "abilities ": [ "EntryAbility " ], 
      "when ": "always " 
    } 
  },
  { 
    "name ": "ohos.permission.INTERNET ", 
    "reason ": "$string:internet ", 
    "usedScene ": { 
      "abilities ": [ "EntryAbility " ], 
      "when ": "always " 
    } 
  } 
],

配置sdk初始化

在您的工程内创建一个AbilityStage类型的组件，如MyAbilityStage.ets，进行sdk的初始化

初始化需要传入sdk的授权文件lic，如果暂时没有，可以传入空，如果需要正式授权，需要联系商务

示例如下

export default class MyAbilityStage extends AbilityStage { 
  onCreate() { 
    let licPath = "rawfile:/sdkLic/com.example.edit.lic" 
    let ctx = meishe.init(this.context, licPath, 0); 
  }
}

初始化人脸模型

初始化人脸模型是拍摄预览人脸特效和编辑预览人脸的基础

首先把模型放到资源目录下,红色框需要用到的模型文件

A71038DE-D045-460D-8C8F-8196E75CB5D6.1.arscene是人脸道具包，可以根据自己的选择添加

初始化人脸相关模型，可以在MyAbilityStage中进行初始化模型

onCreate() { 
  let licPath = "rawfile:/sdkLic/com.example.edit.lic " 
  let ctx = NvsStreamingContext.init(this.context, licPath, 0); 
  let facePath = "rawfile:/face/ms_face240_v3.1.1.next.model " 
  let fakeFacePath = "rawfile:/face/fakeface_v1.0.1.dat " 
  let avatarPath = "rawfile:/face/ms_avatar_v2.0.0.next.model " 
  let handPath = "rawfile:/face/ms_hand_common_v2.0.0.next.model " 
  let humansegPath = "rawfile:/face/ms_humansegment_medium_v2.0.0.next.model " 
  let facecommonPath = "rawfile:/face/facecommon_v1.0.1.dat " 
  let eyecontourPath = "rawfile:/face/ms_eyecontour_v2.0.0.next.model " 
  let advancedbeautyPath = "rawfile:/face/advancedbeauty_v1.0.1.dat " 
  let skysegPath = "rawfile:/face/ms_skysegment_v1.0.5.next.model " 
  this.initHumanDetection(facePath, fakeFacePath, avatarPath, handPath, humansegPath, facecommonPath, eyecontourPath, advancedbeautyPath, skysegPath) 
  Logger.info(TAG,'SDK 初始化完成！ '+ctx); 
}

/**
* 初始化人体检测功能
*
* @param facePath 人脸检测模型路径
* @param fakeFacePath 假脸检测模型路径
* @param avatarPath 虚拟形象表情检测模型路径
* @param handPath 手势检测模型路径
* @param humansegPath 人像分割模型路径
* @param facecommonPath 人脸通用模型路径
* @param eyecontourPath 眼部轮廓检测模型路径
* @param advancedbeautyPath 高级美颜模型路径
* @param skysegPath 天空分割模型路径
* /

initHumanDetection(facePath: string, fakeFacePath: string, avatarPath: string, handPath: string, humansegPath: string, facecommonPath: string, eyecontourPath: string, advan cedbeautyPath: string, skysegPath: string) { 
  let result = NvsStreamingContext.initHumanDetection(facePath, " ", NvsHumanDetectionFeature.keFaceLandmark | NvsHumanDetectionFeature.keFaceAction | NvsHumanDetectionFeature.keSemiImageMode) 
  Logger.info(TAG, `facePath result: ${ result}`) 
  result = NvsStreamingContext.setupHumanDetectionData(NvsHumanDetectionDataType.keFakeFace, fakeFacePath) 
  Logger.info(TAG, `fakeFacePath result: ${result}`) 
  result = NvsStreamingContext.initHumanDetectionExt(avatarPath, " ", NvsHumanDetectionFeature.keAvatarExpression | NvsHumanDetectionFeature.keSemiImageMode) 
  Logger.info(TAG, `avatarPath result: ${result}`) 
  result = NvsStreamingCont ext.initHumanDetectionExt(handPath, " ", NvsHumanDetectionFeature.keHandAction | NvsHumanDetectionFeature.keHandLandmark | NvsHumanDetectionFeature.keSemiImageMode) 
  Logger.info(TAG, `handPath result: ${ result}`) 
  result = NvsStreamingContext.initHumanDetectionExt(humansegPath, " ", NvsHumanDetectionFeature.keSegmentationBackground | NvsHumanDetectionFeature.keSemiImageMode) 
  Logger.info(TAG, `humansegPath result: ${result}`) 
  result = NvsStrea mingContext.setupHumanDetectionData(NvsHumanDetectionDataType.keFaceCommon, facecommonPath) 
  Logger.info(TAG, `facecommonPath result: ${result}`) 
  result = NvsStreamingConte xt.initHumanDetectionExt(eyecontourPath, " ", NvsHumanDetectionFeature.keEyeballLandmark | NvsHumanDetectionFeature.keSemiImageMode) 
  Logger.info(TAG, `eyecontourPath result: ${result}`) 
  result = NvsStreamingC ontext.setupHumanDetectionData(NvsHumanDetectionDataType.keAdvancedBeauty, advancedbeautyPath) 
  Logger.info(TAG, `advancedbeautyPath result: ${result}`) 
  result = NvsStreami ngContext.initHumanDetectionExt(skysegPath, " ", NvsHumanDetectionFeature.keSegmentationSky) 
  Logger.info(TAG, `skysegPath result: ${result}`) 
}

拍摄预览

调用接口之前，需要自行申请摄像头和麦克风的授权，权限申请通过之后再调用sdk的API

在您的工程内创建一个capturePage组件，调用sdk的摄像头预览，示例如下

import meiShe, { NvsLiveWindow, NvsLiveWindowContent, NvsStreamingEngineCaptureFlag, NvsVideoCaptureResolutionGrade} from '@meishe/libnvstreamingsdkcore'

@Builder
export function pageBuilder(name: string, param: Object) { 
  if (name == "capturePage ") { 
    capturePage() 
  } 
}

@Component
struct capturePage { 
  liveWindowContent : NvsLiveWindowContent | null = null; 
  onContentLoad: (content : NvsLiveWindowContent) => void = (content : NvsLiveWindowContent) => { 
    this.liveWindowContent = content; 
  } 
  build() { 
    NavDestination(){ 
      Stack(){ 
        NvsLiveWindow({ onContentLoad: this.onContentLoad }).width( "100% ").height( "100% ") 
          .onAppear(() => { 
            if (this.liveWindowContent) { 
              const suc = meiShe.getInstance()?.connectCapturePreviewWithLiveWindow(this.liveWindowContent) 
              if (suc) { 
                this.startCapturePreview() 
              } 
            } 
          }) 
      }.width('100%').height('100%') 
    }.width('100%').height('100%').title( "拍摄 " ) 
  } 
  
  startCapturePreview() { 
    let flags = NvsStreamingEngineCaptureFlag.NvsStreamingEngineCaptureFlag_StrictPreviewVideoSize | NvsStreamingEngineCaptureFlag.NvsStreamin gEngineCaptureFlag_CaptureBuddyHostVideoFrame | NvsStreamingEngineCaptureFlag.NvsStreamingEngineCaptureFlag_EnableTakePicture | NvsStreamingEngineCaptureFlag.NvsStreamingEn gineCaptureFlag_DontUseSystemRecorder 
    meiShe.getInstance()?.startCapturePreview( 0, NvsVideoCaptureResolutionGrade.NvsVideoCaptureResolutionGradeSupperHigh, f lags, { num: 9, den: 16 } ); 
  } 
}

拍摄使用人脸特技

对拍摄预览画面进行人脸相关特效，例如美颜、美型等

如果使用到了道具效果包，需要先安装

this.arsceneId = Tool.installAssetPackage(ARSCENE_PATH, " ", NvsAssetPackageType.NvsAssetPackageType_ARScene)
if (!this.arSecenFx) { 
  this.arSecenFx = Tool.getSDK()?.appendBuiltinCaptureVideoFx('AR Scene')||null
}
this.arSecenFx?.setNum berVal( "Beauty Whitening ", 1)
this.arSecenFx?.setBooleanVal( "Advanced Beauty Enable ", true)
this.arSecenFx?.setNumberVal( "Advanced Beauty Type ", 0)
this.arSecenFx?.setNumberVal( "Advanced Beauty Intensity ", 1)
this.arSecenFx?.setBooleanVal( "Single Buffer Mode ", false)
/// 人脸遮挡
this.arSecenFx?.setBooleanVal( "AI Face Occlusion Enabled ",true)
this.arSecenFx?.setStringVal('Scene Id',this.arsceneId)

编辑预览

可以在工程里内置一个素材，或者调用系统的相册去获取素材

在您的工程内创建一个editPage组件，拿到素材资源，创建timeline，播放，示例如下

import meiShe, { NvsAudioResolution, NvsBitDepth, NvsLiveWindow, NvsLiveWindowContent, NvsLiveWindowFillMode, NvsTimeline, NvsVideoPreviewSizeMode, NvsVideoResoluti on} from '@meishe/libnvstreamingsdkcore'

@Builder
export function pageBuilder(name: string, param: Object) { 
  if (name == "editPage ") { 
    editPage() 
  } 
}

@Component
struct editPage { 
  videoPath: string = " " 
  liveWindowContent: NvsLiveWindowContent | null = null; 
  onContentLoad: (content: NvsLiveWindowContent) => void = (content: NvsLiveWindowContent) => { 
    this.liveWindowContent = content; 
    this.liveWindowContent.setFillMode(NvsLiveWindowFillMode.NvsLiveWindowFillMode_PreserveAspectFit) 
  } 
  timeline: NvsTimeline | null = null; 
  aboutToAppear(): void { 
    this.timeline = this.createTimeline(); 
  } 
  build() { 
    NavDestination() { 
      Stack() { 
        NvsLiveWindow({ onContentLoad: this.on ContentLoad }).width( "100% ").height( "100% ") 
          .onAppear(() => { 
            if (this.liveWindowContent) { 
              this.createLiveWindow() 
            } 
          }) 
      }.width('100%').height('100%') 
    }.width('100%').height('100% ').title( "拍摄 " ) 
  } 
  
  createTimeline(): NvsTimeline | null { 
    const vr: NvsVideoResolution = { imageWidth: 1920, imageHeight: 1080, bitDepth: NvsBitDepth.NvsBitDepth_8Bit, imagePAR: { num: 1, den: 1 } }; 
    const ar: NvsAudioResolution = { sampleRate: 48000, channelCount: 2 }; 
    const ctx = meiShe.getInstance(); 
    if (ctx != null) { 
      ctx.setDefaultCaptionFade(false) 
      const timeline = ctx?.createTimeline(vr, { num: 30, den: 1 }, ar, 0); 
      if (!timeline) { 
        return null; 
      } 
      const vt rack = timeline?.appendVideoTrack(); 
      vtrack?.appendClip( "rawfile:/test.mp4 " ) 
      return timeline; 
    } else { 
    } 
    return null; 
  } 
  
  createLiveWindow() { 
    if (this.timeline) { 
      const suc = meiShe.getInstance()?.connectTimelineWithLiveWi ndow(this.timeline, this.liveWindowContent) 
      if (suc) { 
      } else { 
      } 
      const seekSuc = meiShe.getInstance()?.playbackTimeline(this.timeline, 0, this.timeline?.ge tDuration(), NvsVideoPreviewSizeMode.NvsVideoPreviewSizeMode_LiveWindowSize, true, 0) 
    } 
  }
}

编辑使用人脸特技

对编辑预览画面进行人脸相关特效，例如美颜、美型等

如果使用到了道具效果包，需要先安装

this.arsceneId = Tool.installAssetPackage(ARSCENE_PATH, " ", NvsAssetPackageType.NvsAssetPackageType_ARScene)
let track = this.timeline?.getVideoTrackByIndex(0)
let clip = track?.getClipByIndex(0)
if (!this.arSecenFx && clip != null) { 
  this.arSecenFx = clip.appendRawBuiltinFx('AR Scene')
}
this.arSecenFx?.getARSceneManipulate().setDetectionMode(32768)
this.arSecenFx?.setNumberVal( "Beauty Whitening ", 1)
this.arSecenFx?.setBooleanVal( "Advanced Beauty Enable ", true)
this.arSecenFx?.setNumberVal( "Advanced Beauty Type ", 0)
this.arSecenFx?.setNumberVal( "Advanced Beauty Intensity ", 1)
this.arSecenFx?.setBooleanVal( "Single Buffer Mode ", false)
/// 人脸遮挡
this.arSecenFx?.setBooleanVal( "AI Face Occlusion Enabled ", true)
this.arSecenFx?.setStringVal('Scene Id', this.arsceneId)

其他功能

参考官网的文档：https://www.meishesdk.com/harmony/doc_ch/html/index.html

技术支持

当出现问题时：

给我们发送邮件：bd@meishesdk.com
联系与您对接的同事，寻求帮助，我们会建立一个支持群为您服务