Tools

Convert buffer offset to milliseconds

Reference Number: AA-01882 Views: 14748

0 Rating/ Voters

The LumenVox API returns the offset for SSML markers and viseme markers in terms of a buffer offset i.e. number of bytes from the 0^th byte. However, that number may not be meaningful without information regarding the sampling rate and audio format. It is sometimes more useful to keep track of the offset in terms of milliseconds.

This page provides some pseudo-code to describe the conversion from buffer offset to number of milliseconds for each of the audio formats we accept. The below pseudo-code is for the C interface, however, the same conversion formula would work for the C++ interface.

C Code

int TTS_handle; // Handle from LV_TTS_CreateClient in C interface
int audio_format, sampling rate; // Got from LumenVox API
int buffer_offset; // Offset returned for ssml or viseme markers
float offset_ms; // Output offset in milliseconds

////////////////////////////////////////////////////////////////////////////
// Insert code that does the synthesis and gets the ssml or viseme
// marker offsets using the LumenVox API
////////////////////////////////////////////////////////////////////////////

// Get audio format and sample rate from TTS Server
LV_TTS_GetSynthesizedAudioFormat(TTSClientHandle, audio_format);
LV_TTS_GetSynthesizedAudioSampleRate(TTS_handle, sampling_rate);

// Calculate milliseconds offset based on audio format
// The below formulas are generic for the C and C++ interfaces
switch(audio_format)
{
case SFMT_ULAW:
case SFMT_ALAW:

// We multiply by 1000 to convert seconds to milliseconds

offset_ms = (float)buffer_offset * 1000 /sampling_rate;

break;

case SFMT_PCM:

// We divide by 2 because there are 2 bytes per PCM sample

offset_ms = (float)buffer_offset / 2 * 1000 / sampling_rate;

break;

}

// Offset_ms now contains buffer_offset converted to milliseconds