Can I clone someone else's voice?

Not without their documented consent, and reputable tools require verification for exactly that reason. Several jurisdictions now protect voice likeness specifically, platforms remove synthetic audio of real people, and the reputational consequence lands on you rather than the vendor. Cloning your own voice for your own content is the straightforward case; everything else needs permission in writing.

How natural does it sound for long-form narration?

Convincing for minutes at a time, with a characteristic flatness that emerges over an audiobook-length listen — the prosody is correct but the performance does not evolve the way a human reader's does. Tools offering SSML or manual control over emphasis, pauses and pace let you fix the worst of it, which matters a great deal more for long content than voice quality does.

Do I need to disclose that a voice is synthetic?

Increasingly yes. Some platforms require labelling of synthetic media, advertising regulators in several markets have taken positions, and audiences react badly to discovering it after the fact. For narration and explainer content disclosure is generally uncontroversial; for anything that could be mistaken for a specific real person speaking, it is essential.