VoxReplica is an AI media studio where students build a “podcast twin” that can speak in their voice without recording every episode. The project starts with ethics-first setup (consent, boundaries, and clear disclosure). Students collect and label a small voice dataset, then train or adapt a voice model for natural prosody, pauses, and emphasis. They also build a narrative engine—episode outlines, templates, fact-check prompts, and tone rules—driving a pipeline from brief → script → synthesis → edit → publish, with optional safeguards like watermarking and speaker verification.