Description

This "text-to-speech"-like voice was originally used in Adult Male #8, American English from TruVoice (pitch 120, speed 160). To compile the dataset, I only used 6 minutes of voicelines from the game. FOR AI COVERS: Keeping the original pitch regularizes the voice from any song. However, transposing it to -12 can improve and correct singing accuracy (to avoid high-pitch notes). Layering octaves from his voice would help, too (which is what I did in the "Ballin'" example. FOR TEXT-TO-SPEECH: If you have any high-pitch audio, transposing it to -12 can correct the voice. I'm thinking about working on a v2 of this model and manually training the original text-to-speech voice even further, avoiding the transposing issue. Feel free to credit me if you want! If you have a request, come DM me and I'll make any AI model for you for free. DL: