the engineer has to locate the voice file in sdk and build the bin file again using a new voice sample in Ubuntu Any firmware developer can take this up work
Success story sharing