I have a Python script that is generating audio with Spark-TTS voice cloning and it's working. However, right now, when I run multiple generations (from the same source voice), the whole process runs for each generation. The source audio and source transcript is loaded, processed, and then an audio file is generated (the cloned voice output). This seems extremely inefficient when I'm running multiple generations with the same source voice. Is there a way that I can load the source audio and source transcript, process it, and then save that result to be used for each of the multiple generations from that same voice? It seems silly that I have to run though duplicate source processing each time. Does anyone have working/tested Python code samples showing how this can be done? Continue reading...
