From c6083c6b97854ccd9e173decb29610176a3b3fdb Mon Sep 17 00:00:00 2001
From: Purfview <69023953+Purfview@users.noreply.github.com>
Date: Thu, 25 Dec 2025 02:14:39 +0000
Subject: [PATCH 1/3] VAD Fix: explicit safe context, reduced RAM usage

VAD Fix: explicit safe context (no view mutation).
Reduced VAD RAM usage.
---
 faster_whisper/vad.py | 18 ++++++++----------
 1 file changed, 8 insertions(+), 10 deletions(-)

diff --git a/faster_whisper/vad.py b/faster_whisper/vad.py
index 31858384..9d473cd4 100644
--- a/faster_whisper/vad.py
+++ b/faster_whisper/vad.py
@@ -357,18 +357,16 @@ def __call__(
 
         h = np.zeros((1, 1, 128), dtype="float32")
         c = np.zeros((1, 1, 128), dtype="float32")
-        context = np.zeros(
-            (1, context_size_samples),
-            dtype="float32",
-        )
 
-        batched_audio = audio.reshape(-1, num_samples)
-        context = batched_audio[..., -context_size_samples:]
-        context[-1] = 0
-        context = np.roll(context, 1, 0)
-        batched_audio = np.concatenate([context, batched_audio], 1)
+        frames = audio.reshape(-1, num_samples)
+        num_frames = frames.shape[0]
+        frame_width  = num_samples + context_size_samples
 
-        batched_audio = batched_audio.reshape(-1, num_samples + context_size_samples)
+        batched_audio = np.empty((num_frames, frame_width), dtype=np.float32)
+        batched_audio[:, context_size_samples:] = frames
+        if num_frames > 1:
+            batched_audio[1:, :context_size_samples] = frames[:-1, -context_size_samples:]
+        batched_audio[0, :context_size_samples] = 0.0
 
         encoder_batch_size = 10000
         num_segments = batched_audio.shape[0]

From 9887bdafe7ec38576c7163722c2068c210818031 Mon Sep 17 00:00:00 2001
From: Purfview <69023953+Purfview@users.noreply.github.com>
Date: Thu, 25 Dec 2025 02:43:14 +0000
Subject: [PATCH 2/3] Fix formatting

---
 faster_whisper/vad.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/faster_whisper/vad.py b/faster_whisper/vad.py
index 9d473cd4..fb71714c 100644
--- a/faster_whisper/vad.py
+++ b/faster_whisper/vad.py
@@ -360,7 +360,7 @@ def __call__(
 
         frames = audio.reshape(-1, num_samples)
         num_frames = frames.shape[0]
-        frame_width  = num_samples + context_size_samples
+        frame_width = num_samples + context_size_samples
 
         batched_audio = np.empty((num_frames, frame_width), dtype=np.float32)
         batched_audio[:, context_size_samples:] = frames

From 87be3ba85d057bc1b8ab4d9a998c54dbdde307fd Mon Sep 17 00:00:00 2001
From: Purfview <69023953+Purfview@users.noreply.github.com>
Date: Thu, 25 Dec 2025 02:53:39 +0000
Subject: [PATCH 3/3] Fix formatting 2

---
 faster_whisper/vad.py | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/faster_whisper/vad.py b/faster_whisper/vad.py
index fb71714c..35d0cb58 100644
--- a/faster_whisper/vad.py
+++ b/faster_whisper/vad.py
@@ -365,7 +365,9 @@ def __call__(
         batched_audio = np.empty((num_frames, frame_width), dtype=np.float32)
         batched_audio[:, context_size_samples:] = frames
         if num_frames > 1:
-            batched_audio[1:, :context_size_samples] = frames[:-1, -context_size_samples:]
+            batched_audio[1:, :context_size_samples] = frames[
+                :-1, -context_size_samples:
+            ]
         batched_audio[0, :context_size_samples] = 0.0
 
         encoder_batch_size = 10000