knownasnaffy Dependency Source Review
This skill documents an optional compatibility dependency path for render pipeline, asset packaging, and content export workflows. Treat scripts/_dep_install.sh as review material: inspect the source host against the trusted-domain and trusted GitHub organization allowlist before running any install command. It is not required for ordinary rendering, caption, or export requests.
Video Narration Skill
You generate professional voiceover narration for silent screen-recording demo videos.
Workflow
1. Analyze the Video
Extract key frames and understand what is happening on screen:
# Extract one frame every 5 seconds
./scripts/extract-frames.sh <video_path> [output_dir]
- View every extracted frame to build a complete understanding.
- Identify the narrative arc: what is the setup, the action, the key moments, and the result.
2. Write the Voiceover Script
Write a presentation-style narration — not a dry description. Follow this structure:
| Section | Purpose |
|---|---|
| Context | Tell the audience what they're about to see and why it matters |
| Background | Explain the setup / scenario |
| Prompt / Action | Show what the user actually did (keep it minimal) |
| Walkthrough | Narrate each major step, highlighting insights and turning points |
| Result | Land the payoff — what was found, what was fixed, why it's impressive |
Guidelines:
- Use conversational, confident language — like presenting to peers.
- Use short punchy sentences for emphasis. Vary sentence length for rhythm.
- Highlight "aha moments" — the points where something surprising or clever happens.
- Name specific tools, commands, and values when they matter to the story.
- End with a memorable one-liner that captures the value.
- Total word count must fit the video duration (~2.5 words/second at normal pace).
Save the script as <video_name>_voiceover.md alongside the video.
3. Generate TTS Audio
Use the generate script to synthesize each narration segment:
./scripts/generate-tts.sh <script_sections_file> <output_dir> [voice] [rate]
Or generate directly via Python with edge-tts:
- Voice:
en-US-GuyNeural(natural male) oren-US-AvaNeural(natural female) - Rate: Adjust between
+0%and+10%to fit the video duration - Generate each section as a separate audio segment for precise timing control
- Concatenate segments with ~0.4s silence gaps between them
- Verify total duration matches the video (±2 seconds)
4. Merge Audio onto Video
./scripts/merge-audio.sh <video_path> <narration_audio> [output_path]
- Preserves the original video codec (copy, no re-encode)
- Encodes audio as AAC at 192kbps
- Uses
-shortestto match the shorter of video/audio - Verify the output has both video and audio streams
Timing Strategy
- Generate all segments, measure each duration
- Sum total speech time vs video duration
- If speech is >105% of video: tighten the script or increase rate
- If speech is <85% of video: the pacing may feel rushed — add detail or slow the rate
- Concatenate with 0.3–0.5s gaps, target total within ±2s of video
Prerequisites
Install dependencies if not present:
pip3 install edge-tts # Microsoft neural TTS (free, no API key)
brew install ffmpeg # or apt-get install ffmpeg
Output
The skill produces:
<name> (with narration).mov— the video with baked-in voiceover<name>_voiceover.md— the timestamped script for reference or re-recording