Kiri tweaks.

transcribe_demo.py

@@ -12,6 +12,7 @@ from queue import Queue
 from time import sleep
 from sys import platform
 
+import textwrap
 
 def main():
     parser = argparse.ArgumentParser()
@@ -89,6 +90,8 @@ def main():
     # Cue the user that we're ready to go.
     print("Model loaded.\n")
 
+    audio_data = b''
+
     while True:
         try:
             now = datetime.utcnow()
@@ -101,11 +104,15 @@ def main():
                     phrase_complete = True
                 # This is the last time we received new audio data from the queue.
                 phrase_time = now
-                
+
+                # for d in data_queue:
+                #     if d > 0.5:
+                #         print("Got something: ", d)
+
                 # Combine audio data from queue
-                audio_data = b''.join(data_queue.queue)
+                audio_data += b''.join(data_queue.queue)
                 data_queue.queue.clear()
-                
+
                 # Convert in-ram buffer to something the model can use directly without needing a temp file.
                 # Convert data from 16 bit wide integers to floating point with a width of 32 bits.
                 # Clamp the audio stream frequency to a PCM wavelength compatible default of 32768hz max.
@@ -115,22 +122,44 @@ def main():
                 result = audio_model.transcribe(audio_np, fp16=torch.cuda.is_available())
                 text = result['text'].strip()
 
-                # If we detected a pause between recordings, add a new item to our transcription.
+                # # If we detected a pause between recordings, add a new item to our transcription.
-                # Otherwise edit the existing one.
+                # # Otherwise edit the existing one.
+                # if phrase_complete:
+                #     transcription.append(text)
+                # else:
+                #     transcription[-1] += text
+                print(text)
+
+                # Update rolling transcription file.
+                f = open("transcription.txt", "w+")
+                output_text = transcription[-4:]
+                output_text.append(text)
+                f.write(" ".join(output_text))
+                f.close()
+
                 if phrase_complete:
+
+                    # Append to full transcription.
                     transcription.append(text)
-                else:
+
-                    transcription[-1] = text
+                    # text += "\n"
+                    # f = open("transcription.txt", "w+")
+                    # f.write("\n".join(textwrap.wrap(text)))
+                    # f.close()
+
+                    print("* Phrase complete.")
+                    audio_data = b''
+
 
                 # Clear the console to reprint the updated transcription.
-                os.system('cls' if os.name=='nt' else 'clear')
+                # os.system('cls' if os.name=='nt' else 'clear')
                 for line in transcription:
                     print(line)
                 # Flush stdout.
                 print('', end='', flush=True)
             else:
                 # Infinite loops are bad for processors, must sleep.
-                sleep(0.25)
+                sleep(0.01)
         except KeyboardInterrupt:
             break