feat: whisper language setting (#134)

* feat: whisper language setting

* fix: set auto as default language setting

src-tauri/src/config.rs

@@ -33,6 +33,8 @@ pub struct Config {
     pub status_check_interval: u64,
     #[serde(skip)]
     pub config_path: String,
+    #[serde(default = "default_whisper_language")]
+    pub whisper_language: String,
 }
 
 #[derive(Deserialize, Serialize, Clone)]
@@ -80,6 +82,10 @@ fn default_status_check_interval() -> u64 {
     30
 }
 
+fn default_whisper_language() -> String {
+    "auto".to_string()
+}
+
 impl Config {
     pub fn load(
         config_path: &PathBuf,
@@ -116,6 +122,7 @@ impl Config {
             auto_generate: default_auto_generate_config(),
             status_check_interval: default_status_check_interval(),
             config_path: config_path.to_str().unwrap().into(),
+            whisper_language: default_whisper_language(),
         };
 
         config.save();
@@ -142,6 +149,12 @@ impl Config {
         self.save();
     }
 
+    #[allow(dead_code)]
+    pub fn set_whisper_language(&mut self, language: &str) {
+        self.whisper_language = language.to_string();
+        self.save();
+    }
+
     pub fn generate_clip_name(&self, params: &ClipRangeParams) -> PathBuf {
         let platform = PlatformType::from_str(&params.platform).unwrap();
 

src-tauri/src/handlers/config.rs

@@ -234,3 +234,14 @@ pub async fn update_status_check_interval(
     state.config.write().await.save();
     Ok(())
 }
+
+#[cfg_attr(feature = "gui", tauri::command)]
+pub async fn update_whisper_language(
+    state: state_type!(),
+    whisper_language: String,
+) -> Result<(), ()> {
+    log::info!("Updating whisper language to {}", whisper_language);
+    state.config.write().await.whisper_language = whisper_language;
+    state.config.write().await.save();
+    Ok(())
+}

src-tauri/src/handlers/video.rs

@@ -400,7 +400,9 @@ pub async fn generate_video_subtitle(
     };
     state.db.add_task(&task).await?;
     log::info!("Create task: {:?}", task);
-    match generate_video_subtitle_inner(&state, &reporter, id).await {
+    let language_hint = state.config.read().await.whisper_language.clone();
+    let language_hint = language_hint.as_str();
+    match generate_video_subtitle_inner(&state, &reporter, id, language_hint).await {
         Ok(result) => {
             reporter.finish(true, "字幕生成完成").await;
             // for local whisper, we need to update the task status to success
@@ -445,6 +447,7 @@ async fn generate_video_subtitle_inner(
     state: &state_type!(),
     reporter: &ProgressReporter,
     id: i64,
+    language_hint: &str,
 ) -> Result<GenerateResult, String> {
     let video = state.db.get_video(id).await?;
     let filepath = Path::new(state.config.read().await.output.as_str()).join(&video.file);
@@ -484,7 +487,9 @@ async fn generate_video_subtitle_inner(
 
                 let mut results = Vec::new();
                 for path in chunk_paths {
-                    let result = generator.generate_subtitle(reporter, &path).await;
+                    let result = generator
+                        .generate_subtitle(reporter, &path, language_hint)
+                        .await;
                     results.push(result);
                 }
 
@@ -548,7 +553,9 @@ async fn generate_video_subtitle_inner(
 
                 let mut results = Vec::new();
                 for path in chunk_paths {
-                    let result = generator.generate_subtitle(reporter, &path).await;
+                    let result = generator
+                        .generate_subtitle(reporter, &path, language_hint)
+                        .await;
                     results.push(result);
                 }
 

src-tauri/src/http_server.rs

@@ -17,8 +17,8 @@ use crate::{
         config::{
             get_config, update_auto_generate, update_clip_name_format, update_notify,
             update_openai_api_endpoint, update_openai_api_key, update_status_check_interval,
-            update_subtitle_generator_type, update_subtitle_setting, update_whisper_model,
-            update_whisper_prompt,
+            update_subtitle_generator_type, update_subtitle_setting, update_whisper_language,
+            update_whisper_model, update_whisper_prompt,
         },
         message::{delete_message, get_messages, read_message},
         recorder::{
@@ -251,6 +251,22 @@ async fn handler_update_whisper_model(
     Ok(Json(ApiResponse::success(())))
 }
 
+#[derive(Debug, Serialize, Deserialize)]
+#[serde(rename_all = "camelCase")]
+struct UpdateWhisperLanguageRequest {
+    whisper_language: Option<String>,
+}
+
+async fn handler_update_whisper_language(
+    state: axum::extract::State<State>,
+    Json(whisper_language): Json<UpdateWhisperLanguageRequest>,
+) -> Result<Json<ApiResponse<()>>, ApiError> {
+    update_whisper_language(state.0, whisper_language.whisper_language)
+        .await
+        .expect("Failed to update whisper language");
+    Ok(Json(ApiResponse::success(())))
+}
+
 #[derive(Debug, Serialize, Deserialize)]
 #[serde(rename_all = "camelCase")]
 struct UpdateSubtitleSettingRequest {
@@ -1227,6 +1243,10 @@ pub async fn start_api_server(state: State) {
             .route(
                 "/api/update_auto_generate",
                 post(handler_update_auto_generate),
+            )
+            .route(
+                "/api/update_whisper_language",
+                post(handler_update_whisper_language),
             );
     } else {
         log::info!("Running in readonly mode, some api routes are disabled");

src-tauri/src/main.rs

@@ -408,6 +408,7 @@ fn setup_invoke_handlers(builder: tauri::Builder<tauri::Wry>) -> tauri::Builder<
         crate::handlers::config::update_openai_api_endpoint,
         crate::handlers::config::update_auto_generate,
         crate::handlers::config::update_status_check_interval,
+        crate::handlers::config::update_whisper_language,
         crate::handlers::message::get_messages,
         crate::handlers::message::read_message,
         crate::handlers::message::delete_message,

src-tauri/src/subtitle_generator.rs

@@ -108,5 +108,6 @@ pub trait SubtitleGenerator {
         &self,
         reporter: &impl ProgressReporterTrait,
         audio_path: &Path,
+        language_hint: &str,
     ) -> Result<GenerateResult, String>;
 }

src-tauri/src/subtitle_generator/whisper_cpp.rs

@@ -38,6 +38,7 @@ impl SubtitleGenerator for WhisperCPP {
         &self,
         reporter: &impl ProgressReporterTrait,
         audio_path: &Path,
+        language_hint: &str,
     ) -> Result<GenerateResult, String> {
         log::info!("Generating subtitle for {:?}", audio_path);
         let start_time = std::time::Instant::now();
@@ -53,8 +54,8 @@ impl SubtitleGenerator for WhisperCPP {
 
         let mut params = FullParams::new(SamplingStrategy::Greedy { best_of: 1 });
 
-        // and set the language to translate to to auto
-        params.set_language(None);
+        // and set the language
+        params.set_language(Some(language_hint));
         params.set_initial_prompt(self.prompt.as_str());
 
         // we also explicitly disable anything that prints to stdout
@@ -179,7 +180,9 @@ mod tests {
             .unwrap();
         let audio_path = Path::new("tests/audio/test.wav");
         let reporter = MockReporter::new();
-        let result = whisper.generate_subtitle(&reporter, audio_path).await;
+        let result = whisper
+            .generate_subtitle(&reporter, audio_path, "auto")
+            .await;
         if let Err(e) = result {
             println!("Error: {}", e);
             panic!("Failed to generate subtitle");

src-tauri/src/subtitle_generator/whisper_online.rs

@@ -56,6 +56,7 @@ impl SubtitleGenerator for WhisperOnline {
         &self,
         reporter: &impl ProgressReporterTrait,
         audio_path: &Path,
+        language_hint: &str,
     ) -> Result<GenerateResult, String> {
         log::info!("Generating subtitle online for {:?}", audio_path);
         let start_time = std::time::Instant::now();
@@ -98,6 +99,8 @@ impl SubtitleGenerator for WhisperOnline {
             .text("response_format", "verbose_json")
             .text("temperature", "0.0");
 
+        form = form.text("language", language_hint.to_string());
+
         if let Some(prompt) = self.prompt.clone() {
             form = form.text("prompt", prompt);
         }
@@ -226,7 +229,11 @@ mod tests {
         assert!(result.is_ok());
         let result = result.unwrap();
         let result = result
-            .generate_subtitle(&MockReporter::new(), Path::new("tests/audio/test.wav"))
+            .generate_subtitle(
+                &MockReporter::new(),
+                Path::new("tests/audio/test.wav"),
+                "auto",
+            )
             .await;
         println!("{:?}", result);
         assert!(result.is_ok());

src/lib/interface.ts

@@ -105,6 +105,7 @@ export interface Config {
   clip_name_format: string;
   auto_generate: AutoGenerateConfig;
   status_check_interval: number;
+  whisper_language: string;
 }
 
 export interface AutoGenerateConfig {

src/page/Setting.svelte

@@ -36,6 +36,7 @@
       encode_danmu: false,
     },
     status_check_interval: 30, // 默认30秒
+    whisper_language: "",
   };
 
   let showModal = false;
@@ -575,6 +576,33 @@
                   </div>
                 </div>
               {/if}
+              <!-- Whisper Language -->
+              <div class="p-4">
+                <div class="flex items-center justify-between">
+                  <div>
+                    <h3
+                      class="text-sm font-medium text-gray-900 dark:text-white"
+                    >
+                      Whisper 语言
+                    </h3>
+                    <p class="text-sm text-gray-500 dark:text-gray-400">
+                      （测试）生成字幕时使用的语言，默认自动识别
+                    </p>
+                  </div>
+                  <div class="flex items-center space-x-2">
+                    <input
+                      type="text"
+                      class="px-3 py-2 bg-gray-100 dark:bg-gray-700 rounded-lg border border-gray-200 dark:border-gray-600 text-gray-900 dark:text-white w-96"
+                      bind:value={setting_model.whisper_language}
+                      on:change={async () => {
+                        await invoke("update_whisper_language", {
+                          whisperLanguage: setting_model.whisper_language,
+                        });
+                      }}
+                    />
+                  </div>
+                </div>
+              </div>
               <div class="p-4">
                 <div class="flex items-center justify-between">
                   <div>