Voice Transcribe
Transcribe audio files using OpenAI's gpt-4o-mini-transcribe model with vocabulary hints and text replacements. Requires uv (https://docs.astral.sh/uv/).
clawhub install voice-transcribeCreate and grow podcasts by planning episodes, producing audio or video, generating clips, and building audience across formats.
# Install Skill (downloads SKILL.md to .claude/skills/) clawhub install podcast # Then just tell Claude: "use Podcast to help me..."
# Same install command — works with all SKILL.md-compatible AI coding tools clawhub install podcast
This Skill is compatible with the OpenClaw standard. After installation, a SKILL.md file is auto-generated, usable by any OpenClaw-compatible AI Agent (Claude Code, Cursor, Windsurf, etc.).
~/podcasts/<show>/
├── brand/ # Cover art, intro/outro, music
├── episodes/ # One folder per episode
│ └── 001/
│ ├── outline.md
│ ├── recording.mp3
│ ├── transcript.md
│ ├── show-notes.md
│ └── clips/
├── guests.md # Guest tracker + relationship notes
└── analytics.md # Performance patterns
formats.md |
| Audio and video production techniques | production.md |
| AI-generated podcast creation | ai-generation.md |
| Growth, SEO, social, monetization | growth.md |
| Episode planning, scripts, show notes | episodes.md |
| Tools, platforms, APIs | tools.md |clawhub install podcast && clawhub podcast plan --topic '科技新闻' --duration 30 && clawhub podcast generate --episode 1 --format audio --output podcast.mp3Transcribe audio files using OpenAI's gpt-4o-mini-transcribe model with vocabulary hints and text replacements. Requires uv (https://docs.astral.sh/uv/).
clawhub install voice-transcribeCreate chapters, highlights, and show notes from podcast audio or transcripts. Use when a user wants chapter markers, highlight clips, or show-note drafts without publishing or distribution actions.
clawhub install podcast-chaptering-highlightsReal cold-start runs of this skill on Claude Code and Codex.
Input: Turn an 8s podcast audio clip into a shareable square 1080×1080 audiogram — an animated waveform synced to the audio over a clean dark background (ffmpeg `showwaves`), keeping BOTH the video and the original audio. The standard podcast-promotion clip format.
Single ffmpeg pass — `showwaves` (cline mode) waveform centered over a dark 0x0E1116 background, mapped with the original audio. Minimal/clean look; no friction.
claude-opus-4-8
Also a single pass — a blue waveform inside a teal-bordered card/panel over the dark background (more designed). Both streams present. Same approach, more visual chrome.
codex-cli 0.140
Output: audiogram.mp4 — 1080×1080 / 8.0s, H.264 video + AAC audio (Claude 980 KB minimal; Codex 809 KB framed). Shown here web-optimized to 720×720.
Verdict: Both produced valid square audiograms (animated waveform synced to audio, both streams) in a single ffmpeg pass — 3 actions each, near-identical speed. Creative split matches the pattern: Claude minimal (bare waveform), Codex more designed (a bordered card). Adds the audio→video-visualization modality + the audio_voice category. The Podcast skill frames the goal ("generate a shareable clip"); ffmpeg’s showwaves executes it.
Environment: macOS · ffmpeg 8.1.1 (showwaves) · symmetric cold start (codex exec vs a fresh Claude Code subagent). The "Podcast" skill is a production/clip-generation workflow skill; the audiogram is its "generate a shareable clip" step, executed via ffmpeg. Test audio: an 8s synthetic clip with amplitude dynamics. No network.