AgentSkillsCN

voice-podcast-kit

借助克隆人声、片头音乐与转场效果，轻松打造多主机播客节目。

SKILL.md
--- frontmatter
name: voice-podcast-kit
description: Create multi-host podcast episodes with cloned voices, intro music, and transitions.
allowed-tools: voice_clone, voice_list, tts, generate_music, list_dir, upload_file

You are running the Voice Podcast Kit skill.

Goal

  • •Produce a complete podcast episode with distinct cloned voices for each host/guest, intro/outro music, and smooth transitions between segments.

Ask for

  • •Episode topic and title.
  • •List of speakers/characters (names and optional voice sample descriptions).
  • •Whether you should clone voices from provided audio samples, or use existing voice IDs.
  • •Episode length target and number of segments (intro, main discussion, listener Q&A, outro).
  • •Any music preferences (genre, mood, tempo).

Workflow

  1. •Confirm speaker lineup and collect voice samples if cloning is requested:
    • •If audio files are provided, call voice_clone for each speaker.
    • •If no samples, call voice_list to show available presets and let user choose.
  2. •Draft a segment-by-segment outline with speaker assignments and timing.
  3. •Generate intro/outro music:
    • •Call generate_music with appropriate mood (upbeat for intro, winding down for outro).
  4. •Write scripts for each segment with clear speaker labels.
  5. •For each spoken segment:
    • •Call tts with the correct voice_id for each speaker.
    • •Use output_format "mp3" for smooth editing.
  6. •Optionally add transition sounds or music beds between segments.
  7. •Return a production清单:
    • •Music files (intro/outro)
    • •Each spoken segment as separate audio files
    • •Full episode script text
    • •Assembly suggestions (timeline order)

Response style

  • •Keep track of which voice_id maps to which speaker name clearly.
  • •Provide file paths organized by segment and type.
  • •Suggest editing workflow but leave final assembly to user.

Notes

  • •Voice consistency across episodes is a key value proposition—encourage users to save voice IDs for recurring hosts.
  • •If the user wants a sample before full production, offer to generate just the intro + first 2 minutes as a preview.