From df087de389546bfabbb8240b15256e46a8e17d6b Mon Sep 17 00:00:00 2001
From: tc-wleite <wladimir.leite@gmail.com>
Date: Sat, 30 Sep 2023 19:11:52 -0300
Subject: [PATCH] '#1909: Use a smaller buffer to avoid Vosk slow down.

---
 .../java/iped/engine/task/transcript/VoskTranscriptTask.java   | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/iped-engine/src/main/java/iped/engine/task/transcript/VoskTranscriptTask.java b/iped-engine/src/main/java/iped/engine/task/transcript/VoskTranscriptTask.java
index bcddd49062..f2aab01193 100644
--- a/iped-engine/src/main/java/iped/engine/task/transcript/VoskTranscriptTask.java
+++ b/iped-engine/src/main/java/iped/engine/task/transcript/VoskTranscriptTask.java
@@ -97,7 +97,8 @@ protected TextAndScore transcribeAudio(File tmpFile) throws Exception {
             int words = 0;
 
             int nbytes;
-            byte[] buf = new byte[1 << 20];
+            // Buffer must be small (see #1909)
+            byte[] buf = new byte[(int) Math.min(tmpFile.length(), 1 << 16)];
             while ((nbytes = ais.read(buf)) >= 0) {
                 if (recognizer.acceptWaveForm(buf, nbytes)) {
                     TextScoreWords result = decodeFromJson(recognizer.getResult());