tencentcloud-sdk-asr 3.0.1179 → 3.0.1192

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA1:
-  metadata.gz: 081a75541335e0863d2746f45309878db8e89298
-  data.tar.gz: 37076a6da3dc5dc65e39b5e251cf5d7b02471251
+  metadata.gz: 03375c81825b28125d30c48da4d062dfbade93b1
+  data.tar.gz: 0f4e074d1912e036ac9315932f72281db60fb6a3
 SHA512:
-  metadata.gz: 2cf4c3f853458d52a95ce7c53c60d1f1b5747d57535add6f198eee10abdb110497d3a6cb50e15e8c289fcb54e07f0116f7f3225a5240d2b64e8797ee2a8fb03f
-  data.tar.gz: 91719b55a4e8e074239f556c115eacba1d001aa9c16634854db0533a0b0c9c6adfcfb58238b2c15d1eecc2a1472cc986bb691a671df0318d2296cdc0cf6d6cef
+  metadata.gz: 51f6f0c4dd0f16ded462b618eb6d692fa496780f2462d6f11f689ea2d0e8ca6b0043590f1dd5f8aeab4ec548b84dcd4f4ad02d11d850bbc0ea82eeed648f109c
+  data.tar.gz: 54ee41a1e97868c9cd0fccb4d64768225dfd54989a984c0b89143b394bc498793b2a20f12dae876428d541a7441658aae3785fa8bf18fe7c01ffb2f1e06c0e98

data/lib/VERSION

@@ -1 +1 @@
-3.0.1179
+3.0.1192

data/lib/v20190614/models.rb

@@ -433,6 +433,7 @@ module TencentCloud
         # @param SpeakerDiarization: 是否开启说话人分离
         # 0：不开启；
         # 1：开启（仅支持以下引擎：8k_zh/8k_zh_large/16k_zh/16k_ms/16k_en/16k_id/16k_zh_large/16k_zh_dialect/16k_zh_en，且ChannelNum=1时可用）；
+        # 3: 开启角色分离，需配合SpeakerRoles参数使用（增值服务，仅支持16k_zh_en引擎，可支持传入声纹对录音文件内的说话人进行角色认证）
         # 默认值为 0
 
         # 注意：
@@ -537,13 +538,21 @@ module TencentCloud
         # 注意：
         # 1. 本功能配置完成后，预计在10分钟后生效
         # @type ReplaceTextId: String
-
-        attr_accessor :EngineModelType, :ChannelNum, :ResTextFormat, :SourceType, :Data, :DataLen, :Url, :CallbackUrl, :SpeakerDiarization, :SpeakerNumber, :HotwordId, :ReinforceHotword, :CustomizationId, :EmotionRecognition, :EmotionalEnergy, :ConvertNumMode, :FilterDirty, :FilterPunc, :FilterModal, :SentenceMaxLength, :Extra, :HotwordList, :KeyWordLibIdList, :ReplaceTextId
+        # @param SpeakerRoles: 开启角色分离能力
+        # 配合SpeakerDiarization: 3 使用，ASR增值服务，可传入一组声纹信息进行角色认证，仅支持16k_zh_en引擎。
+        # 需传入SpeakerRoleInfo数据组，确定说话人的角色信息，涉及RoleAudioUrl和RoleName两个参数。
+        # RoleAudioUrl：需要认证角色的声纹音频地址，建议30s内的纯净人声，最长不能超过45s。
+        # RoleName：需要认证角色的名称，若匹配成功，会替换话者分离中的SpeakerID。
+        # 示例：
+        # "{\"EngineModelType\":\"16k_zh_en\",\"ChannelNum\":1,\"ResTextFormat\":1,\"SourceType\":0,\"Url\":\"需要进行ASR识别的音频链接\",\"SpeakerDiarization\":3,\"SpeakerRoles\":[{\"RoleAudioUrl\":\"需要认证角色的声纹音频地址\",\"RoleName\":\"需要认证角色的名称\"}]}"
+        # @type SpeakerRoles: Array
+
+        attr_accessor :EngineModelType, :ChannelNum, :ResTextFormat, :SourceType, :Data, :DataLen, :Url, :CallbackUrl, :SpeakerDiarization, :SpeakerNumber, :HotwordId, :ReinforceHotword, :CustomizationId, :EmotionRecognition, :EmotionalEnergy, :ConvertNumMode, :FilterDirty, :FilterPunc, :FilterModal, :SentenceMaxLength, :Extra, :HotwordList, :KeyWordLibIdList, :ReplaceTextId, :SpeakerRoles
         extend Gem::Deprecate
         deprecate :ReinforceHotword, :none, 2025, 12
         deprecate :ReinforceHotword=, :none, 2025, 12
 
-        def initialize(enginemodeltype=nil, channelnum=nil, restextformat=nil, sourcetype=nil, data=nil, datalen=nil, url=nil, callbackurl=nil, speakerdiarization=nil, speakernumber=nil, hotwordid=nil, reinforcehotword=nil, customizationid=nil, emotionrecognition=nil, emotionalenergy=nil, convertnummode=nil, filterdirty=nil, filterpunc=nil, filtermodal=nil, sentencemaxlength=nil, extra=nil, hotwordlist=nil, keywordlibidlist=nil, replacetextid=nil)
+        def initialize(enginemodeltype=nil, channelnum=nil, restextformat=nil, sourcetype=nil, data=nil, datalen=nil, url=nil, callbackurl=nil, speakerdiarization=nil, speakernumber=nil, hotwordid=nil, reinforcehotword=nil, customizationid=nil, emotionrecognition=nil, emotionalenergy=nil, convertnummode=nil, filterdirty=nil, filterpunc=nil, filtermodal=nil, sentencemaxlength=nil, extra=nil, hotwordlist=nil, keywordlibidlist=nil, replacetextid=nil, speakerroles=nil)
           @EngineModelType = enginemodeltype
           @ChannelNum = channelnum
           @ResTextFormat = restextformat
@@ -568,6 +577,7 @@ module TencentCloud
           @HotwordList = hotwordlist
           @KeyWordLibIdList = keywordlibidlist
           @ReplaceTextId = replacetextid
+          @SpeakerRoles = speakerroles
         end
 
         def deserialize(params)
@@ -595,6 +605,14 @@ module TencentCloud
           @HotwordList = params['HotwordList']
           @KeyWordLibIdList = params['KeyWordLibIdList']
           @ReplaceTextId = params['ReplaceTextId']
+          unless params['SpeakerRoles'].nil?
+            @SpeakerRoles = []
+            params['SpeakerRoles'].each do |i|
+              speakerroleinfo_tmp = SpeakerRoleInfo.new
+              speakerroleinfo_tmp.deserialize(i)
+              @SpeakerRoles << speakerroleinfo_tmp
+            end
+          end
         end
       end
 
@@ -1635,8 +1653,10 @@ module TencentCloud
         # @type HotwordList: String
         # @param InputSampleRate: 支持pcm格式的8k音频在与引擎采样率不匹配的情况下升采样到16k后识别，能有效提升识别准确率。仅支持：8000。如：传入 8000 ，则pcm音频采样率为8k，当引擎选用16k_zh， 那么该8k采样率的pcm音频可以在16k_zh引擎下正常识别。 注：此参数仅适用于pcm格式音频，不传入值将维持默认状态，即默认调用的引擎采样率等于pcm音频采样率。
         # @type InputSampleRate: Integer
+        # @param ReplaceTextId: 替换词id。用于调用对应的替换词表。
+        # @type ReplaceTextId: String
 
-        attr_accessor :EngSerViceType, :SourceType, :VoiceFormat, :ProjectId, :SubServiceType, :Url, :UsrAudioKey, :Data, :DataLen, :WordInfo, :FilterDirty, :FilterModal, :FilterPunc, :ConvertNumMode, :HotwordId, :CustomizationId, :ReinforceHotword, :HotwordList, :InputSampleRate
+        attr_accessor :EngSerViceType, :SourceType, :VoiceFormat, :ProjectId, :SubServiceType, :Url, :UsrAudioKey, :Data, :DataLen, :WordInfo, :FilterDirty, :FilterModal, :FilterPunc, :ConvertNumMode, :HotwordId, :CustomizationId, :ReinforceHotword, :HotwordList, :InputSampleRate, :ReplaceTextId
         extend Gem::Deprecate
         deprecate :ProjectId, :none, 2025, 12
         deprecate :ProjectId=, :none, 2025, 12
@@ -1647,7 +1667,7 @@ module TencentCloud
         deprecate :ReinforceHotword, :none, 2025, 12
         deprecate :ReinforceHotword=, :none, 2025, 12
 
-        def initialize(engservicetype=nil, sourcetype=nil, voiceformat=nil, projectid=nil, subservicetype=nil, url=nil, usraudiokey=nil, data=nil, datalen=nil, wordinfo=nil, filterdirty=nil, filtermodal=nil, filterpunc=nil, convertnummode=nil, hotwordid=nil, customizationid=nil, reinforcehotword=nil, hotwordlist=nil, inputsamplerate=nil)
+        def initialize(engservicetype=nil, sourcetype=nil, voiceformat=nil, projectid=nil, subservicetype=nil, url=nil, usraudiokey=nil, data=nil, datalen=nil, wordinfo=nil, filterdirty=nil, filtermodal=nil, filterpunc=nil, convertnummode=nil, hotwordid=nil, customizationid=nil, reinforcehotword=nil, hotwordlist=nil, inputsamplerate=nil, replacetextid=nil)
           @EngSerViceType = engservicetype
           @SourceType = sourcetype
           @VoiceFormat = voiceformat
@@ -1667,6 +1687,7 @@ module TencentCloud
           @ReinforceHotword = reinforcehotword
           @HotwordList = hotwordlist
           @InputSampleRate = inputsamplerate
+          @ReplaceTextId = replacetextid
         end
 
         def deserialize(params)
@@ -1689,6 +1710,7 @@ module TencentCloud
           @ReinforceHotword = params['ReinforceHotword']
           @HotwordList = params['HotwordList']
           @InputSampleRate = params['InputSampleRate']
+          @ReplaceTextId = params['ReplaceTextId']
         end
       end
 
@@ -1824,6 +1846,26 @@ module TencentCloud
         end
       end
 
+      # 说话人注册角色声纹信息
+      class SpeakerRoleInfo < TencentCloud::Common::AbstractModel
+        # @param RoleAudioUrl: 音频url地址，建议不超过30秒，最大45秒
+        # @type RoleAudioUrl: String
+        # @param RoleName: 不超过30字节
+        # @type RoleName: String
+
+        attr_accessor :RoleAudioUrl, :RoleName
+
+        def initialize(roleaudiourl=nil, rolename=nil)
+          @RoleAudioUrl = roleaudiourl
+          @RoleName = rolename
+        end
+
+        def deserialize(params)
+          @RoleAudioUrl = params['RoleAudioUrl']
+          @RoleName = params['RoleName']
+        end
+      end
+
       # [录音文件识别](https://cloud.tencent.com/document/product/1093/37823#3.-.E8.BE.93.E5.87.BA.E5.8F.82.E6.95.B0)、[实时语音异步识别](https://cloud.tencent.com/document/product/1093/52061#3.-.E8.BE.93.E5.87.BA.E5.8F.82.E6.95.B0)请求的返回数据
       class Task < TencentCloud::Common::AbstractModel
         # @param TaskId: 任务ID，可通过此ID在轮询接口获取识别状态与结果。TaskId数据类型为**uint64**。

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: tencentcloud-sdk-asr
 version: !ruby/object:Gem::Version
-  version: 3.0.1179
+  version: 3.0.1192
 platform: ruby
 authors:
 - Tencent Cloud
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2025-12-09 00:00:00.000000000 Z
+date: 2025-12-29 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: tencentcloud-sdk-common
@@ -33,9 +33,9 @@ executables: []
 extensions: []
 extra_rdoc_files: []
 files:
-- lib/tencentcloud-sdk-asr.rb
-- lib/v20190614/models.rb
 - lib/v20190614/client.rb
+- lib/v20190614/models.rb
+- lib/tencentcloud-sdk-asr.rb
 - lib/VERSION
 homepage: https://github.com/TencentCloud/tencentcloud-sdk-ruby
 licenses: