Neura Market
prompt logo

prompt

Free

AI prompt for turning long videos into social-ready clips locally

FreeFree tier
Type
Open Source
Visit Website GitHub (40 stars)

About prompt

Social Video Clipify Architect is a structured AI system prompt (intended for models like Claude) that guides an AI assistant to act as a production post-production specialist for turning long-form videos into short, shareable social media clips. It defines a complete workflow using local tools: Whisper for transcription, ffmpeg for video processing, and NumPy for motion-energy analysis. The prompt enforces audio-first discovery (finding funny moments via transcript and waveform), 16:9 to 9:16 reframing with face-pan based on motion energy, and burned-in captions. It outputs clips under 60 seconds optimized for TikTok, Instagram Reels, YouTube Shorts, and LinkedIn vertical video. The pipeline is fully local, requiring no cloud APIs, and includes a user-confirmation step before rendering.

Key Features

Audio-first clip discovery via transcript and waveform analysis
16:9 to 9:16 vertical reframing with face-pan based on motion energy
Local-only toolchain: Whisper, ffmpeg, NumPy — no cloud APIs
Automatic detection of punchlines, reversals, awkward pauses, and quotable moments
Candidate clip proposal (3–5 options) with timestamps and rationale before render
Opus-style captions burned as final overlay step
Supports both English and non-English videos
Targets TikTok, Instagram Reels, YouTube Shorts, LinkedIn vertical video

Pros & Cons

Pros
  • Completely local and private — no video uploads to third-party services
  • Transparent workflow with explicit candidate selection before rendering
  • Uses efficient, lightweight tools (ffmpeg, Whisper tiny.en, NumPy) that run on most machines
  • Captions are optimized for readability and brand consistency
  • Non-English support available by switching Whisper model
Cons
  • Requires technical setup: Python, ffmpeg, and Whisper installation
  • Clip detection relies on heuristic signals (transcript keywords, audio peaks) and may miss nuanced humor
  • Face-pan cropping uses motion energy rather than ML detection, which may be less accurate on crowded scenes
  • User must manually select from proposed clips; no fully automatic output without confirmation
  • Designed for vertical video reframing; not suitable for other aspect ratios without modification

Best For

Repurposing long-form YouTube videos into TikTok/Reels/Shorts clipsExtracting funny moments from live streams or podcasts for social sharingCreating highlight reels from educational or conference recordingsAutomating social media content creation from existing video assetsGenerating vertical video clips for LinkedIn from horizontal presentations

FAQ

What tools are required to use this prompt?
The prompt requires ffmpeg, Whisper (tiny.en or base model), and Python with NumPy. All are installed locally.
Does this prompt use any cloud APIs?
No. The entire pipeline runs on-device with local tools. No cloud SaaS or uploads to external services.
What platforms are the output clips intended for?
The output is optimized for TikTok, Instagram Reels, YouTube Shorts, and LinkedIn vertical video.
How are clip-worthy moments detected?
The prompt uses audio-first discovery: scanning the transcript and waveform for punchlines, reversal moments, awkward pauses, self-roasts, and rapid back-and-forth dialogue.
Does the prompt require user confirmation before rendering?
Yes. The workflow proposes 3–5 candidate clips with timestamps and rationale, and the user must explicitly select which ones to render.