Push-To-Talk should be implemented by toggling the client variable CLIENT_INPUT_DEACTIVATED using the function ts3client_setClientSelfVariableAsInt. The variable can be set to the following values (see the enum InputDeactivationStatus in public_definitions.h):

For Push-To-Talk toggle between INPUT_ACTIVE (talking) and INPUT_DEACTIVATED (not talking).

Example code:

unsigned int error;
bool shouldTalk;

shouldTalk = isPushToTalkButtonPressed();  // Your key detection implementation
if((error = ts3client_setClientSelfVariableAsInt(scHandlerID, CLIENT_INPUT_DEACTIVATED,
                                                 shouldTalk ? INPUT_ACTIVE : INPUT_DEACTIVATED))
    != ERROR_ok) {
    char* errorMsg;
    if(ts3client_getErrorMessage(error, &errorMsg) != ERROR_ok) {
        printf("Error toggling push-to-talk: %s\n", errorMsg);
        ts3client_freeMemory(errorMsg);
    }
    return;
}

if(ts3client_flushClientSelfUpdates(scHandlerID) != ERROR_ok) {
    char* errorMsg;
    if(ts3client_getErrorMessage(error, &errorMsg) != ERROR_ok) {
        printf("Error flushing after toggling push-to-talk: %s\n", errorMsg);
        ts3client_freeMemory(errorMsg);
    }
}

It is not necessary to close and reopen the capture device to implement Push-To-Talk.

Basically it would be possible to toggle CLIENT_INPUT_MUTED as well, but the advantage of CLIENT_INPUT_DEACTIVATED is that the change is not propagated to the server and other connected clients, thus saving network traffic. CLIENT_INPUT_MUTED should instead be used for manually muting the microphone when using Voice Activity Detection instead of Push-To-Talk.

If you need to query the current muted state, use ts3client_getClientSelfVariableAsInt:

int hardwareStatus, deactivated, muted;

if(ts3client_getClientSelfVariableAsInt(scHandlerID, CLIENT_INPUT_HARDWARE,
                                        &hardwareStatus) != ERROR_ok) {
    /* Handle error */
}
if(ts3client_getClientSelfVariableAsInt(scHandlerID, CLIENT_INPUT_DEACTIVATED,
                                        &deactivated) != ERROR_ok) {
    /* Handle error */
}
if(ts3client_getClientSelfVariableAsInt(scHandlerID, CLIENT_INPUT_MUTED,
                                        &muted) != ERROR_ok) {
    /* Handle error */
}

if(hardwareStatus == HARDWAREINPUT_DISABLED) {
    /* No capture device available */
}
if(deactivated == INPUT_DEACTIVATED) {
    /* Input was deactivated for Push-To-Talk (not propagated to server) */
}
if(muted == MUTEINPUT_MUTED) {
    /* Input was muted (propagated to server) */
}

When using Push-To-Talk, you should deactivate Voice Activity Detection in the preprocessor or keep the VAD level very low. To deactivate VAD, use:

ts3client_setPreProcessorConfigValue(serverConnectionHandlerID, "vad", "false");

Output volume

Voice output volume can be adjusted by changing the “volume_modifier” playback option using the function ts3client_setPlaybackConfigValue. The value is in decibel, so 0 is no modification, negative values make the signal quieter and positive values louder.

Example to increate the output volume by 10 decibel:

ts3client_setPlaybackConfigValue(scHandlerID, "volume_modifier", 10);

Input volume

Automatic Gain Control (AGC) takes care of the input volume during preprocessing automatically. Instead of modifying the input volume directory, you modify the AGC preprocessor settings with setProProcessorConfigValue.