podcast-generator

Provide the article or topic and choose a format such as conversational duo or solo explainer. The skill runs parallel web research, then writes a script with speaker tags, segment timing, and production notes at roughly 150 words per minute of audio.

Set the ELEVENLABS_API_KEY environment variable, install elevenlabs and pydub, then render each script line with a mapped voice ID using the eleven_multilingual_v2 model. Concatenate segments with 350ms gaps and export as MP3.

Yes, script-only mode delivers the full script plus production notes including word count, estimated duration, segment breakdown, and voice recommendations. It is the default when no ElevenLabs API key is available.

Yes, use the eleven_multilingual_v2 model for non-English scripts and search for native voices in the target language. Write the script natively in that language and test voices with a short sample before full rendering.

pydub's normalize only adjusts peak levels, not perceived loudness. Use ffmpeg-normalize with two-pass loudnorm targeting -16 LUFS, -1.5 dB true peak, and LRA 11 to meet Apple and Spotify podcast specs.

The multilingual model has a 10K character limit per call and can drift or stutter on very long inputs, so chunk at sentence boundaries under about 800 characters. A 20-minute episode costs roughly $3-5 depending on plan.