/narrate-video

name: narrate-video
metadata:
  version: "1.0.3"
description: Generate professional voiceover narration for a video with audio-video sync using Azure TTS by default, or Gemini 3.1 Flash TTS when configured. Use this skill whenever the user wants to add narration, voiceover, commentary, or voice dubbing to any video file — even if they just say "add audio to this video" or "make a narrated version." Also trigger when the user has a screen recording, demo, tutorial, or presentation video that needs a voice track. Trigger on Chinese requests like "视频配音", "给视频加旁白", "录屏解说", "视频加语音", "视频添加声音", "生成视频旁白", "自动配音", "视频解说词".

Video Narration

Add professional voiceover to a video. Analyze the video, write or refine a timed script, generate speech via Azure TTS or Gemini 3.1 Flash TTS, and merge — producing a narrated video where audio and visuals stay in sync.

**Input**: $ARGUMENTS

Additional resources

• Voice table and timing estimates: [references/voices.md](references/voices.md)
• Gemini TTS API and AI Studio request shape: [references/gemini-tts.md](references/gemini-tts.md)
• Python script template: [scripts/narration_script_template.py](scripts/narration_script_template.py) — copy into the video's directory as `narration_script.py` and fill in the placeholders

---

Phase 0: Setup

Provider

Default to `azure` unless the user explicitly asks for Gemini or already has `GEMINI_API_KEY` configured. When using Gemini, use the official Gemini TTS request pattern documented in [references/gemini-tts.md](references/gemini-tts.md).

Language

Use the language of the actual narration text, not the language of the user's instructions. For Azure, unless the user explicitly chooses a voice:

• English narration: set `NARRATION_LANGUAGE = "en"`; use `en-US-Jimmie:DragonHDFlashLatestNeural`.
• All other narration, including mixed Chinese and English: set `NARRATION_LANGUAGE = "other"`; use `zh-CN-Yunxiao:DragonHDFlashLatestNeural`.

Do not ask for language confirmation when the narration text is already provided. An explicit voice choice overrides these defaults. Gemini voice selection remains provider-specific. See [references/voices.md](references/voices.md) for timing estimates and optional voices.

Environment

Use Python 3.11 or newer. Install or update the stable dependencies in the active virtual environment (`<skill_directory>` is this skill's directory):

python -m pip install --upgrade -r <skill_directory>/requirements.txt

# 1. Check provider credentials exist (NEVER read or display their values)
scripts/check_env.py azure
# or
scripts/check_env.py gemini

# 2. Check tool dependencies
command -v ffmpeg && command -v ffprobe && command -v python3

# 3. Check Python dependencies
python3 -c "import dotenv" 2>&1

# 4. Azure only
python3 -c "import azure.cognitiveservices.speech" 2>&1

If Azure is selected and `AZURE_SPEECH_KEY` or `AZURE_SPEECH_REGION` is missing, ask the user to add them to `~/.narrate_video.env`:

AZURE_SPEECH_KEY=your-key-here
AZURE_SPEECH_REGION=your-region-here

If Gemini is selected and `GEMINI_API_KEY` is missing, ask the user to add it to `~/.narrate_video.env`:

GEMINI_API_KEY=your-key-here
# Optional override
GEMINI_TTS_MODEL=gemini-3.1-flash-tts-preview

Then stop — the key is sensitive, only check whether it exists, never read or display its value.

---

Phase 1: Video Analysis

1.1 Metadata

ffprobe -v quiet -print_format json -show_format -show_streams <video>

Record total duration, resolution, frame rate, and whether an audio track exists.

1.2 Scene extraction

Extract frames at 3–4 second intervals to identify scene transitions:

mkdir -p /tmp/narration-frames
for t in $(seq 0 3 <duration>); do
    ffmpeg -y -ss $t -i <video> -frames:v 1 -q:v 2 /tmp/narration-frames/frame_${t}s.jpg 2>/dev/null
done

Review the frames (use Read tool to view images). For each scene transition, note the precise timestamp. Where timing is ambiguous, extract additional frames at 1–2 second intervals to pinpoint the exact moment.

1.3 Transition map

Build a scene transition table mapping timestamps to visual content:

0s   - Opening screen
3s   - User starts typing
8s   - System begins processing
34s  - Response appears

Narration describing something on screen should start *after* that content is already visible. Viewers notice when audio arrives before the visuals — it feels disorienting. Narrating slightly after the visual appears feels natural, like a presenter walking you through what you're seeing.

---

Phase 2: Script Writing

Format

Each narration segment is a `(start_seconds, text)` tuple:

SEGMENTS = [
    (0, "Opening narration here."),
    (8, "Next segment narration..."),
]

Writing guidance

**Timing**: Leave at least 1 second of silence between segments — this breathing room makes narration feel conversational rather than rushed. Use the timing estimates from [references/voices.md](references/voices.md) to estimate whether text fits: for English, multiply the window (in seconds) by 2.5 words/sec, then take 80% as the safe word count.

**Flow**: Each segment should connect logically to the next. Transition words ("And", "Now", "So") help, but vary them — three consecutive "And now" transitions sound robotic.

**Adapting to input**: If the user provided a draft, calibrate its timestamps against the scene analysis, trim text that overflows its time window, and polish the language — but preserve their intent and key points. Without a draft, write narration for each scene based on what's visible.

**Gemini prompt hygiene**: If using Gemini, keep style instructions separate from the spoken transcript. The script template already wraps text in a safe `TRANSCRIPT:` preamble because Gemini 3.1 Flash TTS can occasionally read metadata aloud or reject vague prompts.

Pre-flight check

Before generating audio, verify each segment fits: