An AI agent skill for generating cinematic videos with AI voiceovers using Microsoft Edge TTS (free, no API key) and Remotion. Covers voice generation, audio-synced scene composition, dynamic duration, and chart/data visualization integration.
This skill was built at the Quantum Athletic Intelligence Lab (QAI Lab) as part of our applied AI toolchain for research communication, product storytelling, and performance science media. At QAI Lab, we believe AI is a force multiplier — a tool that allows researchers, scientists, and practitioners to produce high-quality outputs faster without sacrificing creative control.
The edge-tts-remotion skill was developed to solve a real problem: creating professional video content with voiceovers typically requires expensive TTS APIs, video editing software, and significant production time. This skill eliminates those barriers entirely — using Microsoft Edge's free TTS service and Remotion's React-based video framework to produce cinematic, audio-synced videos from code.
William McDonald Computational Intelligence Scientist — Graduate Student, University of North Carolina Founder | Lead Scientist, Quantum Athletic Intelligence Lab
This skill gives any compatible AI agent the knowledge to:
| Capability | Description |
|---|---|
| Free AI Voiceover | Generate MP3 audio from any text using Microsoft Edge TTS — 323 voices, 70+ languages, no API key |
| Dynamic Video Duration | Auto-size Remotion compositions to match actual audio length via calculateMetadata |
| Scene Composition | Build multi-scene videos with fade/slide/wipe transitions using TransitionSeries |
| Data Visualization | Generate Python/matplotlib charts as SVGs and embed them in video scenes |
| Background Music | Generate ambient music tracks via ffmpeg (no downloads needed) |
| Dual Export | Render to MP4 (H.264) at any resolution via Remotion CLI |
| Tool | Purpose | Cost |
|---|---|---|
| edge-tts | Text-to-speech (323 voices) | Free |
| Remotion | React-based video composition | Free (open source) |
| ffmpeg | Audio generation & processing | Free |
| matplotlib | Chart/data visualization | Free |
| pandoc | Document conversion | Free |
Total API cost: $0
npx skills add The-QAI-Lab/qai-video-skills@edge-tts-remotion -g -ygit clone https://github.com/The-QAI-Lab/qai-video-skills.git
cp -r qai-video-skills/edge-tts-remotion ~/.agents/skills/# edge-tts
pip install edge-tts
# Node.js (v18+) — for Remotion
node --version
# ffmpeg — for audio generation
sudo apt install ffmpeg # Ubuntu/Debian
brew install ffmpeg # macOS
# Python chart generation (optional)
pip install matplotlib numpy# 1. Create project
mkdir my-video && cd my-video
npm init -y
npm install remotion @remotion/cli @remotion/transitions @remotion/media-utils react react-dom typescript @types/react
# 2. Generate voiceovers
npx tsx generate-voiceover.ts
# 3. Preview in browser
npx remotion studio src/index.ts
# 4. Render to MP4
npx remotion render src/index.ts MyComposition out/video.mp4323 voices across 70+ languages and locales. No API key required.
| Voice | Gender | Style |
|---|---|---|
en-US-AriaNeural |
Female | Confident, positive — great for narration |
en-US-GuyNeural |
Male | Passionate — great for promos |
en-US-AndrewNeural |
Male | Warm, authentic — great for storytelling |
en-US-ChristopherNeural |
Male | Authoritative — great for corporate |
en-US-JennyNeural |
Female | Friendly, considerate |
en-GB-SoniaNeural |
Female | British, professional |
en-GB-RyanNeural |
Male | British, professional |
af-ZA-AdriNeural |
Female | Friendly, positive — South African |
See rules/voices.md for the complete list of all 323 voices.
# Slower, lower pitch
edge-tts --voice en-US-GuyNeural --rate=-15% --pitch=-5Hz \
--text "Your text here." --write-media out.mp3
# Faster, energetic
edge-tts --voice en-US-AriaNeural --rate=+20% \
--text "Your text here." --write-media out.mp31. Write scene scripts
↓
2. generate-voiceover.ts → public/voiceover/*.mp3
↓
3. generate-charts.py → public/charts/*.svg (optional)
↓
4. ffmpeg ambient music → public/music/ambient.mp3 (optional)
↓
5. Remotion calculateMetadata reads audio durations → sets composition length
↓
6. npx remotion render → out/video.mp4
This skill works with any agent that reads from ~/.agents/skills/. Verified compatible with:
| Agent | Notes |
|---|---|
| Kiro CLI | Full support — primary development environment |
| Claude Code | Auto-detected on install |
| Codex | Auto-detected on install |
| OpenClaw | Auto-detected on install |
| Cursor Agent | Auto-detected on install |
| Gemini CLI | Auto-detected on install |
| GitHub Copilot | Auto-detected on install |
All agents above are part of the skills.sh ecosystem.
edge-tts-remotion/
├── SKILL.md — skill metadata and rule routing
└── rules/
├── setup.md — project setup, dependencies, folder structure
├── generate-audio.md — edge-tts CLI, batch script, ffmpeg music
├── remotion-integration.md — calculateMetadata, scenes, transitions, audio
└── voices.md — full 323-voice reference table
The QAI Lab intro video was built entirely with this skill:
- 6 scenes with fade transitions
- Voice:
af-ZA-AdriNeuralat -5% rate - Animated neural network grid, HUD overlays, data bars
- 3 Python-generated charts (radar, bar, line)
- Ambient drone music via ffmpeg
- Rendered at 1920×1080, 30fps, ~88 seconds
MIT License — free to use, modify, and distribute.
Quantum Athletic Intelligence Lab combines AI modeling, biomechanics, and data science to deliver precision-driven performance insights. Our research spans computational athlete intelligence, neuro-performance systems, smart sensor integration, and advanced computer vision for human movement analysis.