everything video editing

Video production involves high friction, from manual transcription and jump-cut editing to complex color grading and platform-specific reframing. This skill automates the entire post-production pipeline using a specialized suite of Python tools to turn raw footage into distribution-ready content for social media, ads, or internal projects.

What it does

• Automated editing performs jump cuts, silence removal, and multi-platform reframing with face-aware cropping.
• Audio processing handles Whisper-based transcription, voice isolation via Demucs, music ducking, and automated SFX placement.
• Motion graphics generates brand-aligned lower thirds, call-to-actions, and 3D titles using Remotion and Blender.
• Platform optimization applies safe-zone overlays and export presets for TikTok, Instagram, YouTube, and LinkedIn.
• Creative analysis probes footage for beat detection, scene changes, and loudness normalization to ensure consistent quality.

How it works

1. Footage Analysis: Run analyze.py to generate a technical report and contact sheet of the raw assets.
2. Interactive Brief: The agent asks 1-2 rounds of targeted questions to determine the style, pace, and target platform.
3. Asset Preparation: Scripts generate transcriptions, isolate vocals, and select music beds based on descriptive text.
4. Rendering: A plan.json is constructed and processed by the render engine to produce draft and final exports.

Frameworks & tools

This skill requires FFmpeg and Python. It utilizes Whisper for STT, Demucs for audio separation, Remotion for React-based graphics, Blender for 3D titles, and CLAP for music retrieval.

Why this beats prompting it yourself

Generic prompts cannot handle the granular timing required for jump cuts or the precise coordinate math for safe-zone text placement. This skill provides a structured bridge between natural language intent and the command-line tools necessary for frame-accurate video manipulation.

Use cases

• Social Media Reels: Converting raw talking-head footage into captioned, fast-paced vertical clips with hooks.
• Commercial Ads: Building OVC variants from a script with text supers, brand badges, and voice-overs.
• Content Repurposing: Trimming long-form webinars into highlights with automated silence removal and thumbnails.
• Quick Utility Tasks: Batch compressing files, generating GIFs, or normalizing audio loudness for multiple clips.

Known limitations

Blender 3D titles do not support non-Latin scripts like Bangla. Face tracking utilizes a static crop per clip rather than dynamic frame-by-frame tracking. It does not generate video from text prompts.