Podcast Localization: LLMs for Context-Aware Adaptation

Demonstrating a pipeline that transcribes and diarizes multi‑host English podcasts, translates with cultural nuance to Mandarin, synthesizes matching voices, and outputs synchronized audio via Streamlit.

Overview

I’ll demo a prototype that ingests an English podcast (multi-host), accurately transcribes and diarizes speakers, contextually translates to Mandarin with cultural nuance, synthesizes matching host voices, and reassembles a polished, time-synced audio output—all behind a simple Streamlit UI.

Links

https://coaltlab.notion.site/podcast-localization
AI podcast localization pipeline: WhisperX, GPT-4o, ElevenLabs for speaker-matched translation.

Tech stack