Built with AI

AI-Generated Podcasts
That Sound Human

A methodology and production toolkit for creating multilingual podcasts with synthetic voices — developed by one person using AI-augmented workflows.

Featured Voice Demo — Felix, science narrator
Felix voice demo
0:00
7 Episodes
3 Languages
30+ Voices
300+ Plays / Episode
Top 25% Engagement Globally

Three Voices, One Conversation

AI gets 80% right. The methodology is the other 20% — dialogue dynamics, emotional arcs, and the discipline of knowing when not to let the AI talk.

Host 35%

Guides the listener. Asks questions a curious non-expert would ask. Controls pacing and transitions.

Expert 60%

Delivers the substance. Explains, contextualizes, builds narrative. The voice listeners trust.

Source 5%

The historical or primary voice. Quotes, letters, original words. Adds authenticity and emotional punch.

Emotional Arc Design

Every episode follows a deliberate emotional trajectory — not flat information delivery. Scripts are structured around tension, discovery, and resolution, using audio tags to control vocal delivery at the phrase level.


Five Stages, End to End

Each stage uses AI differently. The skill is knowing which tool for which job — and where human judgment is non-negotiable.

01

Research

Claude

Deep topic research, source analysis, fact compilation. AI reads hundreds of sources and extracts what matters.

02

Script

Claude + Human

Three-voice dialogue writing with emotional arcs, segment pacing, and audio tag markup. The creative core.

03

Generate

ElevenLabs / Chatterbox

Text-to-dialogue generation with full context awareness. Character voices maintain personality across episodes.

04

Post-process

Python + FFmpeg

Automated silence trimming, segment stitching, per-line repair. Custom tools for precision editing.

05

Master

FFmpeg

Loudness normalization to podcast standards (-16 LUFS), music mixing, final export.


30+ Synthetic Voices, Zero Clones

Every voice is designed from scratch — age, accent, personality, speech patterns. The same character sounds like the same person across languages. Voice design is a discipline, not a button click.

Emma

Female, clear English, Dutch accent — bilingual host

HostDutchEnglish
0:00

Felix

Male, warm baritone, wonder narrator — accent-free

NarratorEnglish
0:00

Marc

Male, calm expert, Dutch accent — knowledgeable authority

ExpertDutch
0:00

Zara

Female, energetic and encouraging — accent-free

PresenterEnglish
0:00

Professor

Male, 50+, measured academic — subtle European accent

AcademicEnglish
0:00

Daan

Male, warm news anchor — English with Dutch undertones

AnchorEnglishDutch
0:00