This is really quite good at sounding like Donald, especially for the first half of the audio. I’ll probably play around with this for a bit; it’s. It clear to me how much variation you can get in voice in latent space. Anyway it looks to be a very high quality (at least) short form tts engine with open weights so thanks team!