Cutlery/calibre
1
0
Fork 0
mirror of https://github.com/kovidgoyal/calibre.git synced 2025-07-31 14:33:54 -04:00
Code Issues Packages Projects Releases Wiki Activity

Normalize audio output

Browse Source
This commit is contained in:
Kovid Goyal 2025-07-30 17:10:51 +05:30
parent fa582d8f26
commit 7e45aa6bbd
No known key found for this signature in database
GPG Key ID: 06BC317B515ACE7C
1 changed files with 12 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
src/calibre/utils/tts/piper.cpp
View File

@ -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);