From 93d661f2e4b416770052c28e99a4f2d583259686 Mon Sep 17 00:00:00 2001
From: m-bain <36994049+m-bain@users.noreply.github.com>
Date: Thu, 29 Dec 2022 10:54:23 +0000
Subject: [PATCH] fix whisper hallucination outside of audio length

---
 whisperx/transcribe.py | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/whisperx/transcribe.py b/whisperx/transcribe.py
index 8408ac8..9d9a85e 100644
--- a/whisperx/transcribe.py
+++ b/whisperx/transcribe.py
@@ -292,6 +292,10 @@ def align(
     prev_t2 = 0
     word_segments_list = []
     for idx, segment in enumerate(transcript):
+        if int(segment['start'] * SAMPLE_RATE) >= audio.shape[1]:
+            # original whisper error, transcript is outside of duration of audio, not possible. Skip to next (finish).
+            continue
+        
         t1 = max(segment['start'] - extend_duration, 0)
         t2 = min(segment['end'] + extend_duration, MAX_DURATION)
         if start_from_previous and t1 < prev_t2: