This viewer contains changes designed to improve voice detection and reduce dropouts. It also contains a nice collection of bug and crash fixes.
Voice Activity Detection
This Viewer also exposes four VIVOX VAD (Voice Activity Detection) variables via Debug Settings and disables the (previously enabled) automatic mode. By making changes to these variables, we should be able to come up with a collection of settings that we can base new default values on in settings.xml.
The Debug Settings are:
Enable (1) or disable (0) automatic VAD. If this is disabled, the voice characteristics can be changed by adjusting the other settings. If this is enabled, modifying the other settings will have no effect.
The time (in milliseconds) that it takes for the VAD to switch back to silence from speech mode after the last speech frame has been detected.
A dimensionless value between 0 and 20000 (default 576) that controls the maximum level at which the noise floor may be set at by the VAD’s noise tracking. Too low of a value will make noise tracking ineffective (A value of 0 disables noise tracking and the VAD then relies purely on the sensitivity property). Too high of a value will make long speech classifiable as noise.
A dimensionless value between 0 and 100, indicating the ‘sensitivity of the VAD’. Increasing this value corresponds to decreasing the sensitivity of the VAD (i.e. ‘0’ is most sensitive, while 100 is ‘least sensitive’)
The default values (updated) are (using VIVOX names):
VadHangover(s): 2000 (Valid values are 1 - 60000 milliseconds)
Some of the settings can only be changed by restarting the Viewer or teleporting away and coming back (needs a new voice connection) but the other 2 work in real time as you change them. Early testing suggests that VivoxVadNoiseFloor needs the restart or teleport to stick.
After some initial testing with VIVOX, they suggested starting from a point where VivoxVadSensitivity was set to 0. This will likely result in no dropouts because the microphone is sending everything to the voice channel. However, in a noisy environment (talking in background, vacuum cleaner, TV on etc.) it will also transmit that too. Though with modern microphones, that have built in noise cancellation, sending everything may be a good thing as the microphone may have done all the heavy lifting of noise cancellation first.
Please let us know what VAD Sensitivity value works best for you!
Fernet (Italian pronunciation: [ferˈnɛt]) is an Italian type of amaro, a bitter, aromatic spirit. Fernet is made from a number of herbs and spices which vary according to the brand, but usually include myrrh, rhubarb, chamomile, cardamom, aloe, and especially saffron, with a base of grape distilled spirits.
Fernet is usually served as a digestif after a meal but may also be served with coffee and espresso or mixed into coffee and espresso drinks. It typically contains 45% alcohol by volume. It may be served at room temperature or with ice.
The Italian liqueur has a cult following in the international bartending community and is immensely popular in Argentina. The South American country consumes more than 75% of all fernet produced globally and because of its popularity, it has Fratelli Branca’s only distillery outside of Italy. As it is traditionally mixed with Coke, fernet has also contributed in making Argentina one of the biggest consumers of Coca-Cola in the world. Fernet and Coke (Spanish: fernet con coca) is so ubiquitous in Argentina that it has been described as “the country’s unofficial drink”.
SL-15181 BugSplat Crash: Second Life Release!LLSkinningUtil::initSkinningMatrixPalette(LLMatrix4*, int, LLMeshSkinInfo const*, LLVOAvatar*)
SL-15152 Crash LLError: LLPipeline used when uninitialized