AI Podcast & Audio Newsletter Factory

Fully automated n8n pipeline that runs every morning: ranks top AI stories from RSS feeds, writes a two-host podcast script with Claude, synthesizes each speaker through ElevenLabs with per-persona voice tuning, stitches the audio in-memory, and publishes the episode to S3 and Slack.

n8nAnthropic ClaudeElevenLabs (multi-voice)RSS ingestionAWS S3SlackBinary audio handlingStructured LLM output

Schedule

6am daily, fully unattended

Voices

2 personas, tuned independently

Stitching

In-memory Buffer.concat

Output

MP3 episode → S3 + Slack

Project Overview

Cron-triggered n8n workflow with multi-source RSS ingestion, keyword + recency scoring, a single Claude call for a structured JSON [{speaker, text}] script, per-persona ElevenLabs synthesis (two deliberately differentiated voice profiles), Buffer.concat MP3 stitching in a Code node with no filesystem or ffmpeg dependency, S3 upload, and a Slack episode announcement. Error sub-workflow handles every failure path.

Challenge

Producing a daily audio newsletter manually is hours of work. Generic TTS sounds flat; consistent multi-voice output requires careful per-persona tuning that can't be done ad hoc each run.

Solution

Built a cron-driven n8n pipeline that handles the full production loop — story selection, script generation, per-persona synthesis with codified ElevenLabs settings, in-memory audio stitching via Buffer.concat, and distribution — without any manual intervention.

Results

Fully autonomous daily audio pipeline with two differentiated voice personas, structured LLM output validation, binary audio handling without filesystem access, and per-execution cost tracking.