From e6921a5734c7e46662887ef94de43e0204557b11 Mon Sep 17 00:00:00 2001
From: Raivis Dejus <orvils@gmail.com>
Date: Sat, 7 Mar 2026 05:54:35 +0200
Subject: [PATCH] Adding VAD to whisper.cpp to reduce hallucinations on audio
 with silences

---
 Makefile                                                 | 3 +++
 buzz/transcriber/whisper_cpp.py                          | 5 +++++
 share/metainfo/io.github.chidiwilliams.Buzz.metainfo.xml | 1 +
 3 files changed, 9 insertions(+)

diff --git a/Makefile b/Makefile
index 6a316b53..4beb3323 100644
--- a/Makefile
+++ b/Makefile
@@ -65,6 +65,7 @@ ifeq ($(OS), Windows_NT)
 	cp whisper.cpp/build/bin/Release/whisper-cli.exe buzz/whisper_cpp/
 	cp whisper.cpp/build/bin/Release/whisper-server.exe buzz/whisper_cpp/
 	cp dll_backup/SDL2.dll buzz/whisper_cpp
+	PowerShell -NoProfile -ExecutionPolicy Bypass -Command "if (-not (Test-Path 'buzz\whisper_cpp\ggml-silero-v6.2.0.bin')) { Start-BitsTransfer -Source https://huggingface.co/ggml-org/whisper-vad/resolve/main/ggml-silero-v6.2.0.bin -Destination 'buzz\whisper_cpp\ggml-silero-v6.2.0.bin' }"
 endif
 
 ifeq ($(shell uname -s), Linux)
@@ -82,6 +83,7 @@ ifeq ($(shell uname -s), Linux)
 	cp -P whisper.cpp/build/ggml/src/libggml-base.so* buzz/whisper_cpp/ || true
 	cp -P whisper.cpp/build/ggml/src/libggml-cpu.so* buzz/whisper_cpp/ || true
 	cp -P whisper.cpp/build/ggml/src/ggml-vulkan/libggml-vulkan.so* buzz/whisper_cpp/ || true
+	test -f buzz/whisper_cpp/ggml-silero-v6.2.0.bin || curl -L -o buzz/whisper_cpp/ggml-silero-v6.2.0.bin https://huggingface.co/ggml-org/whisper-vad/resolve/main/ggml-silero-v6.2.0.bin
 endif
 
 # Build on Macs
@@ -101,6 +103,7 @@ endif
 	cp whisper.cpp/build/bin/whisper-server buzz/whisper_cpp/ || true
 	cp whisper.cpp/build/src/libwhisper.dylib buzz/whisper_cpp/ || true
 	cp whisper.cpp/build/ggml/src/libggml* buzz/whisper_cpp/ || true
+	test -f buzz/whisper_cpp/ggml-silero-v6.2.0.bin || curl -L -o buzz/whisper_cpp/ggml-silero-v6.2.0.bin https://huggingface.co/ggml-org/whisper-vad/resolve/main/ggml-silero-v6.2.0.bin
 endif
 
 # Prints all the Mac developer identities used for code signing
diff --git a/buzz/transcriber/whisper_cpp.py b/buzz/transcriber/whisper_cpp.py
index 977b785e..db4aff84 100644
--- a/buzz/transcriber/whisper_cpp.py
+++ b/buzz/transcriber/whisper_cpp.py
@@ -109,6 +109,11 @@ class WhisperCpp:
             "-f", file_to_process,
         ]
     
+        # Add VAD if the model is available
+        vad_model_path = os.path.join(os.path.dirname(whisper_cli_path), "ggml-silero-v6.2.0.bin")
+        if os.path.exists(vad_model_path):
+            cmd.extend(["--vad", "--vad-model", vad_model_path])
+
         # Add translate flag if needed
         if task.transcription_options.task == Task.TRANSLATE:
             cmd.extend(["--translate"])
diff --git a/share/metainfo/io.github.chidiwilliams.Buzz.metainfo.xml b/share/metainfo/io.github.chidiwilliams.Buzz.metainfo.xml
index cc91b618..fd50fb15 100644
--- a/share/metainfo/io.github.chidiwilliams.Buzz.metainfo.xml
+++ b/share/metainfo/io.github.chidiwilliams.Buzz.metainfo.xml
@@ -74,6 +74,7 @@
           <li>Added option to import folder</li>
           <li>Extra settings for live recordings</li>
           <li>Update checker for Windows and Macs</li>
+          <li>Added voice activity detection to whisper.cpp</li>
         </ul>
       </description>
     </release>