Creating a voice clone
I worked with Resemble AI to create a clone of my dad’s voice using six hours of his recordings. When the model of his voice was completed I was able to type words into the interface and hear him speak. Recently I started using Resemble AI’s new Speech-to-Speech tool, which allows a human to speak into the interface to generate synthetic speech that follows the speaker’s inflection, cadence, and pitch.
Researching & writing the script
The process of creating each conversation begins with the collection and transcription of information about my father from different sources including interviews, therapy, a medium and a chatbot. When my research is completed I sift through the transcripts and identify the pieces that resonate with me. I position and rearrange each piece until they start to take the shape of a conversation. The next step is finding the exact words that my father and I will say in the conversation–sometimes the words come directly from the transcript and other times I have to invent the language we would have spoken. The final step is to work with my editor, who is also my sister, to make sure that the content and language of the conversation feels authentic to my father.
Recording the conversation
After I complete the final draft of the script Billy Clark and I record each piece of the conversation. Billy’s recordings are uploaded and transformed into my father’s voice through Resemble AI’s Speech-to-Speech tool. The audio clips are reassembled in ProTools where we adjust the timing and refine the sound quality and then export the conversation as a stereo audio file.