Auto-Clip & Burn AI

Autonomous AI studio turning 30-min videos into viral 9:16 Shorts with Whisper timestamps, speech-density ranking, Fit Canvas mode, and luminous neon karaoke captions in 45s.

Comment

Inspiration

Short-form videos on YouTube Shorts, TikTok, and Instagram Reels are the fastest way to grow an audience today. But turning a long podcast, tutorial, or webinar into vertical Shorts takes a lot of time:

  • Finding the best moments, cropping the video, syncing subtitles, and styling text manually takes over 2 hours per video.
  • Existing tools charge $50 to $80 every month, and they often zoom in too much and cut off important parts of the screen.

I built Auto-Clip & Burn AI as a solo developer to make video repurposing fast, simple, and 100% free. It takes any long horizontal video and automatically creates high-quality 9:16 vertical Shorts with glowing captions in under 45 seconds.

What it does

  • Picks the Best Moments: Analyzes speech pacing and density to automatically find the 3 most exciting, high-energy highlights.
  • Fit Canvas Mode (No Cutoffs): Keeps 100% of your horizontal screen visible in the center with a dynamic blurred background, so code, slides, and text are never cut off.
  • Glowing Subtitles: Uses OpenAI Whisper to sync words with microsecond precision, creating animated neon captions (including Alex Hormozi and MrBeast styles).
  • Hooks & Progress Bar: Automatically adds an attention-grabbing headline banner at the top and an animated progress bar at the bottom.
  • 1-Click Download: Generates crisp 1080x1920 HD MP4 videos ready to download and post immediately.

How we built it

I built the entire project solo using:

  • OpenAI Whisper AI: Transcribes speech locally with word-level microsecond timestamps and fast greedy decoding.
  • FFmpeg 7.x: Handles video cutting, 9:16 vertical canvas framing, blurred background rendering, and subtitle burning.
  • Python & Streamlit: Created a clean dark-mode web studio with live caption previews and creator presets.
  • yt-dlp: Ingests videos directly from YouTube links or local MP4/MOV uploads.

Challenges we ran into

  1. Slow Video Rendering: Rendering high-resolution blurred backgrounds on CPU originally took over 45 seconds per short. I optimized the filter by downscaling the blur layer and enabling multi-threading, which made rendering 10x faster (under 12 seconds).
  2. Cloud YouTube Restrictions: YouTube often blocks cloud servers with 403 errors. I implemented multi-tier client fallbacks with TLS browser impersonation and Node.js challenge solving.
  3. Cross-Platform Compatibility: Replaced Windows-only PowerShell audio synthesis with a universal FFmpeg engine so the 1-click demo runs smoothly on Linux, Mac, and Streamlit Cloud.

Accomplishments that we're proud of

  • 120 mins down to 45 secs: Saves creators over 98% of their post-production time.
  • $0 Infrastructure Cost: Built a complete multi-modal AI studio with zero paid API keys or subscriptions.
  • Fit Canvas Mode: Solved the common problem of traditional clippers cutting off the left and right sides of screen recordings.
  • 100% Solo Delivery: Designed, coded, tested, documented, and deployed the entire app independently.

What we learned

  • Deep understanding of OpenAI Whisper word-level timestamp alignment and greedy decoding speedups.
  • Advanced FFmpeg filtergraph design, complex ASS subtitle styling, and multi-threaded video encoding.
  • How to build fast, responsive creator tools with Streamlit and dark glassmorphic UI principles.

What's next for Auto-Clip & Burn AI

  • Adding automatic B-roll video clip overlays based on spoken keywords.
  • Multi-speaker face tracking that pans smoothly between alternating podcast hosts.
  • 1-click direct publishing to TikTok, YouTube Shorts, and Instagram Reels.

Built With

  • computer-vision
  • ffmpeg
  • moviepy
  • multimodal-ai
  • open-source
  • openai-whisper
  • python
  • pytorch
  • social-media-automation
  • speech-recognition
  • streamlit
  • video-processing
  • yt-dlp
Share this project:

Updates

posted an update —

Auto-Clip & Burn AI v2.3 Release Update

Here is the latest milestone update for Auto-Clip & Burn AI for the Social Media Automation Hackathon:

What is New:

  • Fit Canvas Mode: Displays 100% of your horizontal video with a dynamic blurred background, completely eliminating screen cutoffs and zooming.
  • 10x Faster Rendering: Multi-threaded FFmpeg pipeline renders vertical Shorts in under 12 seconds per clip.
  • 4x Faster Whisper AI: Greedy decoding optimizations for fast speech transcription.
  • Luminous Karaoke Captions: Glowing active-word subtitles with Alex Hormozi, MrBeast, Minimalist, and Cyber style presets.
  • Viral Hooks & Retention Bars: Automatically burns top headline badges and bottom progress bars.

Live Web Studio: https://social-media-autoclipper2.streamlit.app GitHub Codebase: https://github.com/Anurag-tech22/social-media-autoclipper

Log in or sign up for Devpost to join the conversation.

Submission history