Plan L: AI Podcast Production Workflow

ChatGPT / Claude

For interviews write question lists plus guest background; for solo shows write a segmented outline (hook → expansion → action). Give the model your style and revise its first pass.

💡Putting your 3-5 core questions in the outline makes the conversation more natural than a verbatim script.

Live recording / ElevenLabs / MiniMax

With guests, record locally on each side for audio quality; solo, record live or use AI voice. For AI voice use ElevenLabs/MiniMax Chinese voices and audition a few before choosing.

💡For AI dubbing keep sentences short with punctuation; test-record one segment first.

Descript / CapCut

Edit on a text timeline (Descript): delete filler, pauses and repeats by deleting words; fix slip-ups with Overdub. CapCut for segmenting and basic edits.

💡'Edit the text, not the waveform' keeps the feel of speech; listen to deleted spans to confirm the context still flows.

Adobe Podcast / Auphonic

One-click denoise and de-reverb (Adobe Podcast/Auphonic); normalize loudness to podcast standard (~-16 LUFS); add intro/outro BGM kept low.

💡Consistent loudness reads more 'professional' than raw sound quality; keep BGM under 30% of voice.

Podcast hosting platform

Create cover art, shownotes (AI-generated summary + timestamps + links) and RSS; publish to XiaoYuzhou, Apple Podcasts and Spotify; build a collection page.

💡Timestamps in shownotes let listeners jump straight to what they want — higher completion rate.