From 0180b1f29e5f0a81c7dadfde04f09a49c903ac86 Mon Sep 17 00:00:00 2001
From: Adolfo Reyna <aeroreyna@gmail.com>
Date: Thu, 26 Feb 2026 20:53:39 -0500
Subject: [PATCH] Initial commit: basic Whisper live transcription script

---
 .gitignore                             |   1 +
 __pycache__/transcribe.cpython-310.pyc | Bin 0 -> 1589 bytes
 transcribe.py                          |  66 +++++++++++++++++++++++++
 3 files changed, 67 insertions(+)
 create mode 100644 .gitignore
 create mode 100644 __pycache__/transcribe.cpython-310.pyc
 create mode 100644 transcribe.py

diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000..8b73e14
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1 @@
+__pycache__/\n*.pyc\n.DS_Store
diff --git a/__pycache__/transcribe.cpython-310.pyc b/__pycache__/transcribe.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..3ddde216e0949b9d27dc0ca731bf825175884b9c
GIT binary patch
literal 1589
zcmZux&2QT_6elT4mgSE)ZQ8WJu<6jDGqxG%3gl1>Mbpf~f+kKE2Pn#mC!36;M3z*N
zYAVAiT4dK?_Z0=4<NhUGcgktkVTWapa#Qp$68t@q@A2`y$H$j68Xkhy{d*q1tswNL
zS$x_EEFQs7hX5RL93i%#n2~aAF<Mfa*(G(D!wIW!i@BV#Dz}-(9aa<eync>YgIBow
zD`HJv<sQ%_(c-luBJYc(Q-nc(zZ~uQ1>A>v(~1t`yd&ZZ49oD%Lx4wLAar5u8c`$j
zp_#cE8!*%zfSDz$8c^dy!toiwNLzC{Lv#BaonyEP^--Jj+PI+mVI<m4VW%>Tbz!Mo
z6^_zeNLkc^jEBkKMJ6)AtTfXF9VKbrrpyUr9_WBM`!bjaRZtyHggF=JO#Ojm9c=ig
zp=A)Okqn2Tljemx3ZiHjjE-8yN(X@ACccJc6Q&~=ADiKsmnq#@1LrlwMdNpfllRss
zn6JqkTsy}nhO^i~oR%EK?UK?%dyF(abhvXyPVjVBR}NiW)t;{DI<K73-yULAMY=I>
zo+Gsmbm<(O5KZT;8J>~(G6V|FE%@@DzBa?B7|vf`K}WVcn4!a!88+|8YiHJpHM4mA
zHk#i69dt(JFJM<=M&F_L<o~RB@$EgNqGRyH^ye_1l}|#EzL~#W?1F}-Is5#Ivzq^o
z(TR=Gv<bYWnSG`GSt)a3zlI!y3<8<g_Ax$2?N)EPx|IYxjK}`7gHWYH`jdo<$iF+K
zcRQWVw6R?-BM=h2GhOrgI*^ccV?PRC3jgW>QynHT$opSCl|m_hL(Ayv4PPg|(n;E$
z?(`BZ9{4-4KM6)V{crp{$zYoMu@KydjUf}Xza5Tbk{%?n@WGTxs1&GGh7a=6*Jdvc
zqK?<b@Sk*^3P?3i1(P%qGSGs#qk|xhMWo7{DXPOL86Bx`D)xRXh=|*S(f#z^TH!`P
zJkEl#fW#E9^mJ+OLnfAvJn*N>;AoPjVB9eGGbLna>Uv!%6Zybz-;y>YU*T=<{LtMR
z>^*<lEh@*0@G;MXYETBTAaTm<K*}Izm3`RP_rEKM;;gnA!(ZuZDS}Di_1CwbZgqFp
z_qv6<@zZ***WKzb%=aE|?QA?5^fy^oy5>e(Oj4Z})>!Dm3BbIV8_j{%A}*{5U}<l6
zz1QE^-F(~~7?sSsdZM8yY^7zG7RygWK1>41H)DvX%u-!cyRSx~^u6scyR@Fw7oU5W
z?e7a&Q1hJ&2VOUew7h1l!J_=e3uu@fItjw~2F%nNKnr`=#V&Diix9Gc$zqP(e_M9T
zs$)u;xMh*QZeG?`n()&<!|oe$DZezD+`+);8VuUYas@V+9cPm?mqt0OsU(Xz7caw+
zkT;FEU6vqe>Vmv(;Ff_^1Gf!)4#4c6O`R$uS;qI{O)-?H!Qc~TOm&i_NAC9!j4TpR
b3&1NMz(wFw*{MbGqJxjbwcR@&rkDQ#TZ)-T

literal 0
HcmV?d00001

diff --git a/transcribe.py b/transcribe.py
new file mode 100644
index 0000000..f3ddd85
--- /dev/null
+++ b/transcribe.py
@@ -0,0 +1,66 @@
+import whisper
+import numpy as np
+import sounddevice as sd
+import queue
+import sys
+
+# Parameters
+MODEL_TYPE = "tiny.en"
+CHANNELS = 1
+SAMPLERATE = 16000
+BLOCK_SIZE = 8000 # 0.5 seconds of audio per block
+TRANSCRIBE_RATE = 2 # Process every 2 seconds
+
+audio_queue = queue.Queue()
+
+def callback(indata, frames, time, status):
+    if status:
+        print(status, file=sys.stderr)
+    audio_queue.put(indata.copy())
+
+def main():
+    print(f"Loading Whisper model '{MODEL_TYPE}'...")
+    model = whisper.load_model(MODEL_TYPE)
+    print("Model loaded.")
+
+    print("\nAvailable Audio Devices:")
+    devices = sd.query_devices()
+    print(devices)
+    
+    # Try to find a sensible default if the system one is tricky
+    default_device = sd.default.device[0]
+    print(f"\nUsing default input device index: {default_device}")
+
+    print("\nStarting live transcription... (Press Ctrl+C to stop)")
+    print("Note: On macOS, you may need to grant Microphone permissions to your terminal.\n")
+    
+    audio_buffer = np.array([], dtype=np.float32)
+
+    try:
+        with sd.InputStream(samplerate=SAMPLERATE, channels=CHANNELS, callback=callback, blocksize=BLOCK_SIZE):
+            while True:
+                # Pull all available data from the queue
+                while not audio_queue.empty():
+                    data = audio_queue.get()
+                    audio_buffer = np.append(audio_buffer, data.flatten())
+
+                # If we have enough audio, transcribe it
+                if len(audio_buffer) >= SAMPLERATE * TRANSCRIBE_RATE:
+                    # Transcribe the current buffer
+                    # fp16=False is used for CPU execution
+                    result = model.transcribe(audio_buffer, fp16=False, language="en")
+                    text = result['text'].strip()
+                    
+                    if text:
+                        print(f"Transcription: {text}")
+                    
+                    # Clear buffer for next chunk
+                    audio_buffer = np.array([], dtype=np.float32)
+
+    except KeyboardInterrupt:
+        print("\nStopped by user.")
+    except Exception as e:
+        print(f"\nError: {e}")
+
+if __name__ == "__main__":
+    main()