Merge pull request #10 from biu7/main

update paraformer model v2
devinyf · Nov 12, 2024 · 9e84fae · 9e84fae
2 parents 91ec842 + 3df837b
commit 9e84fae
Show file tree

Hide file tree

Showing 3 changed files with 15 additions and 4 deletions.
diff --git a/example/paraformer/voice_file/recordfile2text.go b/example/paraformer/voice_file/recordfile2text.go
@@ -12,7 +12,7 @@ import (
 )
 
 func main() {
-	model := paraformer.ParaformerV1
+	model := paraformer.ParaformerV2
 	token := os.Getenv("DASHSCOPE_API_KEY")
 	if token == "" {
 		panic("token is empty")
@@ -29,12 +29,14 @@ func main() {
 	filePath := "file://" + voiceFile
 
 	req := &paraformer.AsyncTaskRequest{
-		Model: paraformer.ParaformerV1,
+		Model: paraformer.ParaformerV2,
 		Input: paraformer.AsyncInput{
 			// 官方示例中使用的远程文件.
 			// FileURLs: []string{"https://dashscope.oss-cn-beijing.aliyuncs.com/samples/audio/paraformer/hello_world_female2.wav"},
 			// 本地文件.
-			FileURLs: []string{filePath},
+			FileURLs:                 []string{filePath},
+			DisfluencyRemovalEnabled: true,
+			LanguageHints:            []string{"zh", "en"},
 		},
 		Download: true, // 是否下载异步任务结果.
 	}

diff --git a/paraformer/dtypes.go b/paraformer/dtypes.go
@@ -76,7 +76,11 @@ type AsyncTaskRequest struct {
 }
 
 type AsyncInput struct {
-	FileURLs []string `json:"file_urls"`
+	FileURLs                 []string `json:"file_urls"`
+	VocabularyID             string   `json:"vocabulary_id,omitempty"`              // 热词表 ID.
+	ChannelID                []int    `json:"channel_id,omitempty"`                 // 音轨索引.
+	DisfluencyRemovalEnabled bool     `json:"disfluency_removal_enabled,omitempty"` // 过滤语气词.
+	LanguageHints            []string `json:"language_hints,omitempty"`             // 识别语音中语言的代码列表. 仅对paraformer-v2生效
 }
 
 type AsyncTaskResponse struct {

diff --git a/paraformer/params.go b/paraformer/params.go
@@ -9,9 +9,14 @@ const (
 	ParaformerV1    ModelParaformer = "paraformer-v1"
 	Paraformer8KV1  ModelParaformer = "paraformer-8k-v1"
 	ParaformerMtlV1 ModelParaformer = "paraformer-mtl-v1"
+	ParaformerV2    ModelParaformer = "paraformer-v2"
+	Paraformer8KV2  ModelParaformer = "paraformer-8k-v2"
+
 	// real time voice.
 	ParaformerRealTimeV1   ModelParaformer = "paraformer-realtime-v1"
 	ParaformerRealTime8KV1 ModelParaformer = "paraformer-realtime-8k-v1"
+	ParaformerRealTimeV2   ModelParaformer = "paraformer-realtime-v2"
+	ParaformerRealTime8KV2 ModelParaformer = "paraformer-realtime-8k-v2"
 )
 
 const (