Merge pull request #126 from solaoi/main

update to v0.9.9

.github/workflows/release.yml

@@ -82,3 +82,8 @@ jobs:
         releaseDraft: true
         prerelease: false
         args: --target aarch64-apple-darwin
+
+    - name: Contribute List
+      uses: akhilmhdh/contributors-readme-action@v2.3.6
+      env:
+        GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

README.md

@@ -44,8 +44,8 @@ unzip ./src-tauri/resources/vosk-model-ja-$VOSK_MODEL_VERSION.zip -d ./src-tauri
 rm ./src-tauri/resources/vosk-model-ja-$VOSK_MODEL_VERSION.zip
 
 # whisper
-curl -LO https://huggingface.co/datasets/ggerganov/whisper.cpp/resolve/main/ggml-large.bin
-mv ggml-large.bin ./src-tauri/resources
+curl -LO https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-large-v1.bin
+mv ggml-large-v1.bin ./src-tauri/resources/ggml-large.bin
 ```
 
 ### voskのライブラリ更新
@@ -77,3 +77,8 @@ yarn tauri dev
 ```
 yarn tauri build
 ```
+
+## Contributors
+
+<!-- readme: contributors -start -->
+<!-- readme: contributors -end -->

package.json

@@ -1,7 +1,7 @@
 {
   "name": "lycoris",
   "private": true,
-  "version": "0.9.8",
+  "version": "0.9.9",
   "type": "module",
   "license": "MIT",
   "engines": {

src-tauri/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "lycoris"
-version = "0.9.8"
+version = "0.9.9"
 description = "Lycoris is an offline voice memo"
 authors = ["solaoi"]
 license = "MIT"

src-tauri/migrations/001.sql

@@ -42,6 +42,7 @@ INSERT INTO models(model_name, model_type) VALUES("base", "whisper");
 INSERT INTO models(model_name, model_type) VALUES("base.en", "whisper");
 INSERT INTO models(model_name, model_type) VALUES("large", "whisper");
 INSERT INTO models(model_name, model_type) VALUES("large-distil.en", "whisper");
+INSERT INTO models(model_name, model_type) VALUES("large-distil.ja", "whisper");
 INSERT INTO models(model_name, model_type) VALUES("medium", "whisper");
 INSERT INTO models(model_name, model_type) VALUES("medium.en", "whisper");
 INSERT INTO models(model_name, model_type) VALUES("small", "whisper");

src-tauri/src/module/model_type_whisper.rs

@@ -5,6 +5,7 @@ pub enum ModelTypeWhisper {
     BaseEn,
     Large,
     LargeDistilEn,
+    LargeDistilJa,
     Medium,
     MediumEn,
     Small,
@@ -20,6 +21,7 @@ impl ModelTypeWhisper {
             ModelTypeWhisper::BaseEn => "base.en",
             ModelTypeWhisper::Large => "large",
             ModelTypeWhisper::LargeDistilEn => "large-distil.en",
+            ModelTypeWhisper::LargeDistilJa => "large-distil.ja",
             ModelTypeWhisper::Medium => "medium",
             ModelTypeWhisper::MediumEn => "medium.en",
             ModelTypeWhisper::Small => "small",
@@ -39,6 +41,7 @@ impl FromStr for ModelTypeWhisper {
             "base.en" => Ok(ModelTypeWhisper::BaseEn),
             "large" => Ok(ModelTypeWhisper::Large),
             "large-distil.en" => Ok(ModelTypeWhisper::LargeDistilEn),
+            "large-distil.ja" => Ok(ModelTypeWhisper::LargeDistilJa),
             "medium" => Ok(ModelTypeWhisper::Medium),
             "medium.en" => Ok(ModelTypeWhisper::MediumEn),
             "small" => Ok(ModelTypeWhisper::Small),

src-tauri/src/module/record.rs

@@ -131,6 +131,10 @@ impl Record {
         .expect("Could not build stream");
 
         stream.play().expect("Could not play stream");
+
+        let app_handle = self.app_handle.clone();
+        app_handle.clone().emit_all("readyToRecognize", "").unwrap();
+
         let (stop_writer_tx, stop_writer_rx) = sync_channel(1);
         let is_converting = Arc::new(Mutex::new(false));
         let (stop_convert_tx, stop_convert_rx) = unbounded();

src-tauri/src/module/record_desktop.rs

@@ -160,6 +160,9 @@ impl RecordDesktop {
         );
         stream.start_capture().ok();
 
+        let app_handle = self.app_handle.clone();
+        app_handle.clone().emit_all("readyToRecognize", "").unwrap();
+
         let (stop_writer_tx, stop_writer_rx) = sync_channel(1);
         let is_converting = Arc::new(Mutex::new(false));
         let (stop_convert_tx, stop_convert_rx) = unbounded();

src-tauri/src/module/transcriber.rs

@@ -12,8 +12,10 @@ impl Transcriber {
             model_type = "small";
         } else if transcription_accuracy.starts_with("medium") {
             model_type = "medium"
-        } else if transcription_accuracy.starts_with("large-distil.en"){
+        } else if transcription_accuracy.starts_with("large-distil.en") {
             model_type = "large-distil.en"
+        } else if transcription_accuracy.starts_with("large-distil.ja") {
+            model_type = "large-distil.ja"
         } else if transcription_accuracy.starts_with("large") {
             model_type = "large"
         }
@@ -24,7 +26,8 @@ impl Transcriber {
             .to_string_lossy()
             .to_string();
 
-        return WhisperContext::new_with_params(&model_path, WhisperContextParameters::default()).expect("failed to load whisper model");
+        return WhisperContext::new_with_params(&model_path, WhisperContextParameters::default())
+            .expect("failed to load whisper model");
     }
 
     pub fn build_params(

src-tauri/tauri.conf.json

@@ -8,7 +8,7 @@
   },
   "package": {
     "productName": "Lycoris",
-    "version": "0.9.8"
+    "version": "0.9.9"
   },
   "tauri": {
     "allowlist": {

src/components/molecules/RecordPreparingButton.tsx

@@ -0,0 +1,10 @@
+const RecordPreparingButton = (): JSX.Element => {
+    return (
+        <button className="btn gap-2 glass text-secondary" disabled>
+            <div className="radial-progress animate-spin-slow" style={{ "--value": 85, "--size": "1.4rem", "--thickness": "2px" } as React.CSSProperties} />
+            録音準備中
+        </button>
+    )
+}
+
+export { RecordPreparingButton }

src/components/molecules/TranscriptionAccuracy.tsx

@@ -43,6 +43,12 @@ const TranscriptionAccuracy = (): JSX.Element => {
                 return "文字起こし：高";
             case "large-distil.en":
                 return "文字起こし：英";
+            case "large-distil.ja":
+                return "文字起こし：日";
+            case "online-transcript":
+                return "文字起こし：オンライン";
+            case "online-chat":
+                return "AI：オンライン";
             case "small-translate-to-en":
                 return "翻訳（英）：低";
             case "medium-translate-to-en":
@@ -102,6 +108,13 @@ const TranscriptionAccuracy = (): JSX.Element => {
                                     return a
                                 }
                             }
+                            if (c === "large-distil.ja") {
+                                if (speakerLanguage?.startsWith("ja") || speakerLanguage?.startsWith("small-ja")) {
+                                    return [...a, c]
+                                } else {
+                                    return a
+                                }
+                            }
                             return [...a, c, `${c}-translate-to-en`]
                         }, []).map((model, i) => (
                             <li key={"transcription-accuracy_" + i}>

src/components/organisms/NoteMain.tsx

@@ -18,6 +18,7 @@ import { TraceStartButton } from '../molecules/TraceStartButton'
 import { TraceStopButton } from '../molecules/TraceStopButton'
 import { tracingState } from '../../store/atoms/tracingState'
 import { tracingNoteState } from '../../store/atoms/tracingNoteState'
+import { RecordPreparingButton } from '../molecules/RecordPreparingButton'
 
 const NoteMain = (): JSX.Element => {
     const transcriptionAccuracy = useRecoilValue(transcriptionAccuracyState)
@@ -42,6 +43,7 @@ const NoteMain = (): JSX.Element => {
             setShowGotoBottom(true);
         }
     }, []);
+    const [isReadyToRecognize, setIsReadyToRecognize] = useState(false);
     useEffect(() => {
         const scrollContainer = scrollContainerRef.current;
         if (scrollContainer) {
@@ -121,6 +123,19 @@ const NoteMain = (): JSX.Element => {
         }
     }, [recordingNote, isTracing])
 
+    useEffect(() => {
+        if (isRecording) {
+            const unlistenReadyToRecognize = listen('readyToRecognize', () => {
+                setIsReadyToRecognize(true);
+            });
+            return () => {
+                unlistenReadyToRecognize.then(f => f());
+            }
+        } else {
+            setIsReadyToRecognize(false);
+        }
+    }, [isRecording])
+
     return (<>
         <div className="max-w-7xl mx-auto py-2 px-4 sm:px-6 lg:px-8 bg-white flex items-center group relative overflow-x-hidden" style={{ height: "64px" }} >
             <h1 className="overflow-hidden select-none text-ellipsis whitespace-nowrap text-2xl tracking-tight font-bold text-gray-600 flex-1 cursor-pointer mr-2 hover:border-base-300 border-2 border-transparent rounded-lg"
@@ -152,7 +167,7 @@ const NoteMain = (): JSX.Element => {
                     <TraceStartButton />}
             </div>
             <div className="flex-none">
-                {(isRecording && recordingNote === selectedNote?.note_id) ? <RecordStopButton /> : <RecordStartButton />}
+                {(isRecording && recordingNote === selectedNote?.note_id) ? isReadyToRecognize ? <RecordStopButton /> : <RecordPreparingButton /> : <RecordStartButton />}
             </div>
             <div className={`absolute top-0 -inset-full h-full w-1/2 z-5 block transform -skew-x-12 bg-gradient-to-r from-transparent to-red-100 opacity-40 ${(isRecording && recordingNote === selectedNote?.note_id) && "animate-shine"}`} />
             <div className={`absolute top-0 -inset-full h-full w-1/2 z-5 block transform -skew-x-12 bg-gradient-to-r from-transparent to-yellow-100 opacity-40 ${(isTracing && tracingNote === selectedNote?.note_id) && "animate-shine"}`} />

src/components/organisms/SettingsMain.tsx

@@ -330,6 +330,13 @@ const SettingsMain = (): JSX.Element => {
                         </div>
                         <ModelDownloadWhisperProgress modelType="large-distil.en" />
                     </div>
+                    <div style={{ height: "86px" }}>
+                        <div className="flex items-center justify-between mb-2">
+                            <p>言語パック（日本語：速度優先）</p>
+                            <ModelDownloadWhisperButton modelType="large-distil.ja" />
+                        </div>
+                        <ModelDownloadWhisperProgress modelType="large-distil.ja" />
+                    </div>
                 </div>
             </div>
             <div className="px-5">