Merge pull request #43 from silvioprog/detected-spoken-language

feat: detected spoken language (fix #40)
JacobLinCool · Aug 28, 2024 · 028208a · 028208a
2 parents 2987b73 + b898a98
commit 028208a
Show file tree

Hide file tree

Showing 3 changed files with 13 additions and 0 deletions.
diff --git a/.changeset/slow-poems-retire.md b/.changeset/slow-poems-retire.md
@@ -0,0 +1,5 @@
+---
+"smart-whisper": minor
+---
+
+Detected spoken language
diff --git a/src/binding/transcribe.cc b/src/binding/transcribe.cc
@@ -248,6 +248,9 @@ class TranscribeWorker : public Napi::AsyncProgressQueueWorker<int> {
                 confidence = confidence / (tokens - skips);
             }
 
+            segment.Set(
+                "lang",
+                Napi::String::New(Env(), whisper_lang_str(whisper_full_lang_id_from_state(state))));
             segment.Set("confidence", Napi::Number::New(Env(), confidence));
             segment.Set("tokens", tokens_array);
         }
@@ -304,6 +307,9 @@ class TranscribeWorker : public Napi::AsyncProgressQueueWorker<int> {
                     confidence = confidence / (tokens - skips);
                 }
 
+                segment.Set("lang",
+                            Napi::String::New(
+                                Env(), whisper_lang_str(whisper_full_lang_id_from_state(state))));
                 segment.Set("confidence", Napi::Number::New(Env(), confidence));
                 segment.Set("tokens", tokens_array);
             }

diff --git a/src/types.ts b/src/types.ts
@@ -77,6 +77,8 @@ export interface TranscribeSimpleResult {
  */
 export interface TranscribeDetailedResult<TokenTimestamp extends boolean>
 	extends TranscribeSimpleResult {
+	/** The detected spoken language. */
+	lang: string;
 	/** The confidence level of the transcription, calculated by the average probability of the tokens. */
 	confidence: number;
 	/** The tokens generated during the transcription process. */