Add Qt application

.vscode/settings.json

@@ -0,0 +1,5 @@
+{
+  "files.associations": {
+    "Buzz.spec": "python"
+  }
+}

Buzz.spec

@@ -5,7 +5,6 @@ datas = []
 datas += collect_data_files('torch')
 datas += copy_metadata('tqdm')
 datas += copy_metadata('torch')
-datas += copy_metadata('tqdm')
 datas += copy_metadata('regex')
 datas += copy_metadata('requests')
 datas += copy_metadata('packaging')
@@ -23,8 +22,8 @@ a = Analysis(
     pathex=[],
     binaries=[],
     datas=datas,
-    hiddenimports=['googleapiclient', 'apiclient', 'pytorch', '“sklearn.utils._cython_blas”',
-                   '“sklearn.neighbors.typedefs”', '“sklearn.neighbors.quad_tree”', '“sklearn.tree”', '“sklearn.tree._utils”'],
+    hiddenimports=['apiclient', 'pytorch', '“sklearn.utils._cython_blas”', '“sklearn.neighbors.typedefs”',
+                   '“sklearn.neighbors.quad_tree”', '“sklearn.tree”', '“sklearn.tree._utils”'],
     hookspath=[],
     hooksconfig={},
     runtime_hooks=[],

main.py

@@ -1,9 +1,82 @@
+import logging
 import warnings
 
+from PyQt5.QtCore import *
+from PyQt5.QtWidgets import *
+
 from transcriber import Transcriber
 
-# logging.basicConfig(level=logging.DEBUG)
+logging.basicConfig(level=logging.DEBUG)
 warnings.filterwarnings('ignore')
 
-transcriber = Transcriber()
-transcriber.start_recording()
+
+class TranscriberWorker(QObject):
+    text = pyqtSignal(str)
+    finished = pyqtSignal()
+
+    def __init__(self, *args) -> None:
+        super().__init__(*args)
+        self.transcriber = Transcriber(text_callback=self.on_next_text)
+
+    def run(self):
+        self.transcriber.start_recording()
+        self.finished.emit()
+
+    def on_next_text(self, text: str):
+        self.text.emit(text)
+
+    def stop_recording(self):
+        self.transcriber.stop_recording()
+
+
+class Application:
+    def __init__(self) -> None:
+        self.app = QApplication([])
+        self.window = QWidget()
+
+        layout = QVBoxLayout()
+
+        record_button = QPushButton("Record")
+        record_button.clicked.connect(self.on_click_record)
+
+        stop_button = QPushButton("Stop")
+        stop_button.clicked.connect(self.on_click_stop)
+
+        self.text_box = QTextEdit()
+        self.text_box.setReadOnly(True)
+
+        layout.addWidget(record_button)
+        layout.addWidget(stop_button)
+        layout.addWidget(self.text_box)
+
+        self.window.setLayout(layout)
+
+    def on_next_text(self, text: str):
+        self.text_box.append(text)
+
+    def on_click_record(self):
+        self.thread = QThread()
+
+        self.transcriber_worker = TranscriberWorker()
+        self.transcriber_worker.moveToThread(self.thread)
+
+        self.thread.started.connect(self.transcriber_worker.run)
+        self.transcriber_worker.finished.connect(self.thread.quit)
+        self.transcriber_worker.finished.connect(
+            self.transcriber_worker.deleteLater)
+        self.thread.finished.connect(self.thread.deleteLater)
+
+        self.transcriber_worker.text.connect(self.on_next_text)
+
+        self.thread.start()
+
+    def on_click_stop(self):
+        self.transcriber_worker.stop_recording()
+
+    def start(self):
+        self.window.show()
+        self.app.exec()
+
+
+app = Application()
+app.start()

transcriber.py

@@ -2,17 +2,20 @@ import logging
 import os
 import wave
 from datetime import datetime
+from typing import Callable
 
 import pyaudio
 import whisper
 
 
 class Transcriber:
-    def __init__(self, model_name="tiny") -> None:
+    def __init__(self, model_name="tiny", text_callback: Callable[[str], None] = print) -> None:
         self.pyaudio = pyaudio.PyAudio()
         self.model = whisper.load_model(model_name)
         self.stream = None
         self.frames = []
+        self.text_callback = text_callback
+        self.stopped = False
 
     def start_recording(self, frames_per_buffer=1024, sample_format=pyaudio.paInt16, channels=1, rate=44100, chunk_duration=4):
         logging.debug("Recording...")
@@ -26,9 +29,14 @@ class Transcriber:
         self.stream.start_stream()
 
         frames_per_chunk = int(rate / frames_per_buffer * chunk_duration)
-
         while True:
+            if self.stopped:
+                self.frames = []
+                logging.debug("Recording stopped. Exiting...")
+                return
             if len(self.frames) > frames_per_chunk:
+                logging.debug("Buffer size: %d. Transcribing next %d frames..." %
+                              (len(self.frames), frames_per_chunk))
                 chunk_path = self.chunk_path()
                 try:
                     clip = []
@@ -36,18 +44,19 @@ class Transcriber:
                         clip.append(self.frames[i])
                     frames = b''.join(clip)
 
+                    # TODO: Can we pass the chunk to whisper in-memory?
                     self.write_chunk(chunk_path, channels, rate, frames)
 
                     result = self.model.transcribe(
                         audio=chunk_path, language="en")
 
-                    # TODO: this should probably be a callback or output buffer
-                    print(result["text"])
+                    logging.debug("Received next result: \"%s\"" %
+                                  result["text"])
+                    self.text_callback(result["text"])
 
                     os.remove(chunk_path)
 
                     self.frames = self.frames[frames_per_chunk:]
-                    logging.debug("Buffer size: ", len(self.frames))
                 except KeyboardInterrupt as e:
                     self.stop_recording()
                     os.remove(chunk_path)
@@ -59,6 +68,7 @@ class Transcriber:
 
     def stop_recording(self):
         logging.debug("Ending recording...")
+        self.stopped = True
         self.stream.stop_stream()
         self.stream.close()
         self.pyaudio.terminate()