video fast

Generating narrated Chinese videos is often slowed down by fragmented documentation and repetitive boilerplate. This skill provides a token-lean, automated pipeline for creating 30-second to 3-minute explainer videos, promo clips, and scientific shorts using HyperFrames. It replaces heavy manual workflows with condensed contracts and pre-built scripts for narration, subtitles, and music.

What it does

• Automated composition converts data-driven story schemas into complete scripts, storyboards, and animation frames.
• Narrative-first audio generates TTS voiceovers with word-level timestamps, automated breathing gaps, and background music synthesis.
• Natural subtitles breaks captions by grammatical clauses rather than word density to ensure readability on mobile and desktop.
• Visual consistency provides a kit of motion-graphic primitives for hooks, feature lists, and code demos with automatic cue-syncing.
• Automated QA runs a "preflight" environment check and a "look" tool to detect layout clipping, timing mismatches, and rendering inconsistencies.

How it works

1. Environment Preflight: Run preflight.py to verify ffmpeg, edge-tts, and font dependencies before starting work.
2. Scripting: Define the message and duration in SCRIPT.md, choosing voice samples and marking delivery tones.
3. Audio Generation: Execute tts_lines.py to create the narration track, which serves as the master clock for the entire project.
4. Visual Assembly: Use story_build.py for data-driven layout or new_frame.py for custom GSAP-based animation frames.
5. Validation & Render: Run look.mjs to catch errors like text clipping or empty frames, then render the final MP4.

Frameworks & tools

HyperFrames, GSAP (GreenSock), FFmpeg, edge-tts, whisper-cli, and Python-based automation scripts.

Why this beats prompting it yourself

Prompting a video agent from scratch usually leads to timing drifts, poor Chinese sentence breaks, and high token costs from reading massive framework docs. This skill uses a specialized contract and local scripts to reduce token input by 10x while ensuring frame-perfect synchronization between voice and motion.

Use cases

• Software demos using the panel-demo prototype to show terminal commands and UI steps.
• Knowledge sharing videos with clause-based subtitles for high-density information.
• Short-form promos for social media using kinetic typography and card-row layouts.
• Workflow automation for re-rendering existing videos with new voices or updated captions.

Known limitations

Frames are hardcoded to 16:9 by default; vertical 9:16 layouts require manual band-top adjustments for captions. Parallel sub-agent execution is not supported on standard pi-web hosts.