I want to build a focused AI-driven audio transformation system whose single job is to modulate voice pitch for podcast episodes. The goal is to let me take spoken-word recordings and shift the pitch up or down in a natural, artifact-free way so hosts or guests can sound more consistent across segments. Here is the scope I have in mind: • Core engine: a machine-learning or DSP solution—built in Python with libraries such as PyTorch, TensorFlow, or a lightweight C++/Rust backend if latency demands it—that reads standard WAV/MP3 input, applies the pitch shift, and exports the processed file at the same sample rate. • Control parameters: semitone or cents sliders plus a formant-preservation toggle so the speaker still sounds authentic. • Workflow fit for podcasting: batch processing of multiple tracks, optional real-time preview, and metadata passthrough so chapter markers stay intact. • Simple UI: either a minimal desktop app (Electron, Qt, or similar) or a CLI with clear flags; I’m flexible as long as it’s easy to integrate into my existing Adobe Audition and Audacity pipeline. Acceptance criteria: 1. A demo file proving transparent pitch alteration with no noticeable artifacts. 2. Source code with installation instructions. 3. Brief documentation explaining how to adjust parameters for different vocal ranges. If you have previous experience in voice modulation, audio DSP, or podcast post-production tooling, I’d love to see examples when you respond.