create-podcast

Podcast Generator (Create Podcast)

Skill Overview

Automatically convert daily briefings, weekly newsletters, articles, or any text into podcast audio. Supports monologue-style narration using your cloned voice and generates all the assets needed for one-click publishing on Xiaoyuzhou.

Use Cases

1. Converting Daily Briefings to Audio

Automatically fetch the daily briefing, select key topics, and generate a 10–12-minute news podcast. Suitable for content creators who want to expand text-based briefings to audio channels, or users who prefer receiving daily news by listening.

2. Converting Weekly Newsletters/Articles into Podcasts

Convert weekly tech newsletters, blog articles, or documents into podcast audio. Suitable for knowledge creators who want to distribute content across multiple channels, or readers who want to explore long-form content through audio.

3. Automated Podcast Publishing

Automatically complete the entire production process for a Xiaoyuzhou podcast episode, from script generation and audio synthesis to cover design. Suitable for creators who want to produce podcast content in batches or personal brands that need a consistent podcast output.

Core Features

1. Multi-Format Content Conversion

Supports converting various content sources into podcasts: daily briefings (selecting 2–4 items for in-depth coverage from 10 pieces of content), weekly newsletters (linking articles by category), individual articles (extracting key points), and arbitrary Markdown/text. Scripts are automatically divided into segments, supporting high-quality synthesis for 10–20-minute audio.

2. Fish.audio TTS Integration

Uses the Fish.audio API to synthesize speech with your cloned voice, with adjustable speaking speed (0.5–2.0). Automatically divides long text into segments to avoid quality degradation, supports retrying individual failed segments, and uses FFmpeg to automatically merge the audio and normalize its volume.

3. Xiaoyuzhou Publishing Asset Generation

After audio synthesis is complete, automatically generates the full set of publishing assets: episode titles with keywords (such as “EP12 · Claude 4/Cursor New Features · 03.14 Daily Briefing”), ShowNotes with timestamp anchors, and a 1:1 square cover image. These can be copied directly into the Xiaoyuzhou creator dashboard for publishing.

Frequently Asked Questions

How can I automatically convert a daily briefing into a podcast?

Read the briefing file generated by bestblogs-daily-digest, extract the titles, summaries, and links of 10 items, and retrieve the full text of the top three in-depth pieces. Generate a 10–12-minute podcast script based on the daily briefing script template. After confirmation, automatically call Fish.audio TTS to synthesize the audio, then generate the publishing assets for Xiaoyuzhou.

What configuration is required to generate a podcast with my own voice?

Before first use, complete voice cloning through Fish.audio and obtain FISH_AUDIO_VOICE_ID. You can configure EXTEND.md to set the speaking speed (1.0 by default) and the name used in the introduction. Once the FISH_AUDIO_API_KEY and FISH_AUDIO_VOICE_ID environment variables are configured, you can narrate podcasts using your own voice.

How do I publish a generated podcast on Xiaoyuzhou?

After audio synthesis is complete, the system automatically generates the assets required by Xiaoyuzhou: an episode title (no more than 50 Chinese characters, including keywords and the date), ShowNotes (in Markdown format, including timestamp anchors and original links), and a cover image (1:1 square, 3000×3000). Copy podcast.mp3, the cover image, and the ShowNotes content into the Xiaoyuzhou creator dashboard to publish.

Is the podcast script generated automatically, or does it need to be written manually?

A first draft is automatically generated using script templates for different content types: daily briefings, weekly newsletters, individual articles, and free-form content each have their own template. The script style is anchored to contents/style-profile.md, which injects your personal voice characteristics. After generation, the user must confirm or revise the script to ensure content quality and TTS compatibility (plain text, conversational language, and no special symbols).

What content formats can be converted into podcasts?

Supported input formats include daily briefings (digest.txt), weekly newsletters (retrieved through the BestBlogs API), article URLs (retrieved through WebFetch), local files (Markdown/TXT), and directly pasted text. All output is standardized as MP3 audio, accompanied by metadata.json, a cover image, and ShowNotes.

What are the usage limitations of the Fish.audio API?

A valid FISH_AUDIO_API_KEY and FISH_AUDIO_VOICE_ID are required. Common errors include: 401, indicating an invalid API key; 404, indicating that the Voice ID does not exist; and 429, indicating a rate limit (wait 10 seconds before retrying). Failed segments are automatically retried once; if they still fail, they are marked and the system continues processing the remaining segments.

How can I process multiple articles in batches to generate podcasts?

The workflow supports reusing Workspace cache (contents/tmp/workspace/YYYY-MM-DD/article-details/). Previously retrieved article details can be reused directly to avoid duplicate API calls. Multiple articles can be processed in parallel, and failed audio segments can be retried individually. The final output can consist of separate episodes or be merged into a podcast collection.

How is the podcast RSS feed updated automatically?

The podcast-rss.ts script can add metadata for new episodes to the RSS feed, supporting automatic synchronization through subscriptions on platforms such as Apple Podcasts, Spotify, and Google Podcasts. R2_PUBLIC_URL must be configured as the base URL for the audio files. Xiaoyuzhou supports direct uploads and does not depend on RSS.