From 6f60d49c7b5103e871817afda0638fd0affe2fea Mon Sep 17 00:00:00 2001
From: Adolfo Reyna <aeroreyna@gmail.com>
Date: Sat, 28 Feb 2026 19:00:56 -0500
Subject: [PATCH] feat: implement streaming mode (-s) and 4-bit quantization
 (-q) for performance

---
 README.md     |  2 ++
 transcribe.py | 22 ++++++++++++++++++++++
 2 files changed, 24 insertions(+)

diff --git a/README.md b/README.md
index 32c134d..2cd2512 100644
--- a/README.md
+++ b/README.md
@@ -55,6 +55,8 @@ Run the script using `python3 transcribe.py` with optional flags.
 - `-i`, `--ingest`: Enable data transmission to the remote server.
 - `-l`, `--list-devices`: Show available audio devices and exit.
 - `-d`, `--device [ID]`: Input device index (bypasses selection prompt).
+- `-q`, `--quantize`: Use 4-bit quantized Whisper model for faster transcription (Strategy 3).
+- `-s`, `--stream`: Enable real-time streaming transcription/draft mode (Strategy 1).
 
 ---
 
diff --git a/transcribe.py b/transcribe.py
index 78f6468..d42876d 100644
--- a/transcribe.py
+++ b/transcribe.py
@@ -83,6 +83,7 @@ def ingest_worker():
         ingest_queue.task_done()
 
 def main():
+    global WHISPER_MODEL
     parser = argparse.ArgumentParser(description="Live transcription and translation with Whisper.")
     parser.add_argument("-es", action="store_true", help="Enable Spanish translation")
     parser.add_argument("-en", action="store_true", help="Enable English (detected or bridged)")
@@ -91,9 +92,14 @@ def main():
     parser.add_argument("-i", "--ingest", action="store_true", help="Enable data transmission to server")
     parser.add_argument("-l", "--list-devices", action="store_true", help="Show available audio devices and exit")
     parser.add_argument("-d", "--device", type=int, help="Input device index")
+    parser.add_argument("-q", "--quantize", action="store_true", help="Use 4-bit quantized Whisper model for speed")
+    parser.add_argument("-s", "--stream", action="store_true", help="Enable real-time streaming transcription (Draft mode)")
     
     args = parser.parse_args()
 
+    if args.quantize:
+        WHISPER_MODEL = "mlx-community/whisper-small-mlx-4bit"
+
     if args.list_devices:
         print("\nAvailable Audio Devices:")
         print(sd.query_devices())
@@ -140,6 +146,7 @@ def main():
     
     audio_buffer = []
     speech_started = False
+    last_stream_time = time.time()
 
     try:
         with sd.InputStream(samplerate=SAMPLERATE, channels=CHANNELS, callback=callback, blocksize=BLOCK_SIZE, device=device_index):
@@ -167,6 +174,11 @@ def main():
                         
                         if (buffer_len_samples - last_end) > (SAMPLERATE * MIN_SILENCE_DURATION_MS / 1000) or buffer_len_samples > BUFFER_LIMIT:
                             
+                            # Clear draft line if it was used
+                            if args.stream:
+                                sys.stdout.write("\r\033[K")
+                                sys.stdout.flush()
+
                             # 1. Transcribe & Detect Language
                             transcription_result = mlx_whisper.transcribe(current_audio, path_or_hf_repo=WHISPER_MODEL)
                             original_text = transcription_result['text'].strip()
@@ -220,6 +232,16 @@ def main():
                             
                             audio_buffer = []
                             speech_started = False
+                            last_stream_time = time.time()
+
+                        elif args.stream and (time.time() - last_stream_time) > 1.0:
+                            # Draft transcription
+                            draft_result = mlx_whisper.transcribe(current_audio, path_or_hf_repo=WHISPER_MODEL)
+                            draft_text = draft_result['text'].strip()
+                            if draft_text:
+                                sys.stdout.write(f"\r\033[K[DRAFT]: {draft_text}")
+                                sys.stdout.flush()
+                            last_stream_time = time.time()
                     
                     elif not speech_started and len(current_audio) > SAMPLERATE * 2:
                         audio_buffer = []