From 7e45aa6bbdb7cd3fa3c3f3d2875fae3626fee4ed Mon Sep 17 00:00:00 2001
From: Kovid Goyal <kovid@kovidgoyal.net>
Date: Wed, 30 Jul 2025 17:10:51 +0530
Subject: [PATCH] Normalize audio output

---
 src/calibre/utils/tts/piper.cpp | 13 ++++++++++++-
 1 file changed, 12 insertions(+), 1 deletion(-)

diff --git a/src/calibre/utils/tts/piper.cpp b/src/calibre/utils/tts/piper.cpp
index 02d73dc77b..85e797f320 100644
--- a/src/calibre/utils/tts/piper.cpp
+++ b/src/calibre/utils/tts/piper.cpp
@@ -440,13 +440,23 @@ next(PyObject *self, PyObject *args) {
     PyObject *ans = NULL, *data = NULL;
     int num_of_silence_samples = 0;
     if (current_sentence_delay > 0) num_of_silence_samples = (int)(current_sample_rate * current_sentence_delay);
+    const float *normalized = audio_tensor_data;
+    if (num_samples) {
+        float maxval = std::abs(audio_tensor_data[0]), q;
+        for (int i = 1; i < num_samples; i++) if ((q = std::abs(audio_tensor_data[i])) > maxval) maxval = q;
+        if (maxval > 1e-8) {
+            float *temp = (float*)malloc(num_samples * sizeof(audio_tensor_data[0]));
+            for (int i = 0; i < num_samples; i++) temp[i] /= maxval;
+            normalized = temp;
+        }
+    }
     if (as_16bit_samples) {
         data = PyBytes_FromStringAndSize(NULL, sizeof(int16_t) * (num_samples + num_of_silence_samples));
         if (data) {
             Py_BEGIN_ALLOW_THREADS;
             int16_t *x = (int16_t*)PyBytes_AS_STRING(data);
             for (int i = 0; i < num_samples; i++) {
-                x[i] = (int16_t)(std::max(-1.f, std::min(audio_tensor_data[i], 1.f)) * std::numeric_limits<int16_t>::max());
+                x[i] = (int16_t)(audio_tensor_data[i]  * std::numeric_limits<int16_t>::max());
             }
             memset(x + num_samples, 0, num_of_silence_samples * sizeof(int16_t));
             Py_END_ALLOW_THREADS;
@@ -461,6 +471,7 @@ next(PyObject *self, PyObject *args) {
             Py_END_ALLOW_THREADS;
         }
     }
+    if (normalized != audio_tensor_data) free((void*)normalized);
     if (data) {
         ans = Py_BuildValue(
             "OiiO", data, num_samples, current_sample_rate, phoneme_id_queue.empty() ? Py_True : Py_False);