Workflow: 1) Normalize show notes and add SSML for pacing. 2) Pick one ElevenLabs voice_id and fixed settings; save them. 3) Batch in groups of 5–10, run 2–4 concurrent jobs, and implement exponential backoff + retry on 429s (check ElevenLabs docs for exact limits). 4) Post-process: loudness to -16 LUFS, trim silence, encode at fixed bitrate. 5) File naming S01_E01_Title_v1.mp3 and write ID3 tags (title, episode, date, description) with ffmpeg or mutagen.