diff --git a/aegisub/audio_provider_lavc.cpp b/aegisub/audio_provider_lavc.cpp
index a3a8e4f7d..a46941e43 100644
--- a/aegisub/audio_provider_lavc.cpp
+++ b/aegisub/audio_provider_lavc.cpp
@@ -196,14 +196,16 @@ void LAVCAudioProvider::GetAudio(void *buf, int64_t start, int64_t count)
 					we should do something about it, not pretend that everything's OK. -Fluff */
 					decoded_samples = audio_resample(rsct, _buf, buffer, decoded_samples / codecContext->channels);
 
-					/* make sure we somehow didn't end up with more samples than we wanted */
-					assert(decoded_samples <= samples_to_decode);
+					/* make some noise if we somehow ended up with more samples than we wanted (will cause audio skew) */
+					if (decoded_samples <= samples_to_decode)
+						wxLogMessage(wxString::Format(_T("Warning: decoder output more samples than requested, audio skew highly likely! (Wanted %d, got %d)"), samples_to_decode, decoded_samples));
 				} else {
 					/* no resampling needed, just copy to the buffer */
 					/* if (decoded_samples > samples_to_decode)
 						decoded_samples = samples_to_decode; */
-					/* I do not understand the point of the above, changed to a more reasonable assertation instead -Fluff */
-					assert(decoded_samples <= samples_to_decode);
+					/* I do not understand the point of the above -Fluff */
+					if (decoded_samples <= samples_to_decode)
+						wxLogMessage(wxString::Format(_T("Warning: decoder output more samples than requested, audio skew highly likely! (Wanted %d, got %d)"), samples_to_decode, decoded_samples));
 
 					memcpy(_buf, buffer, temp_output_buffer_size);
 				}