Voice Cloning with Emotion

There's no preset emotion list to pick from — the clone's baseline comes from how the sample was delivered, and delivery controls adjust pacing and tone from there.

Generally yes, since the clone learns from the delivery in the sample, so a more expressive recording gives the model more expression to work with.

Talking naturally, in the tone you actually want reflected, tends to work better than a stiff scripted read, because that tone is what carries through.

Generate a short line and listen — that tells you more than any description, since results vary with the sample and the text you generate.