Vijay Kumar 👋
A Polyglot Software Developer 💻 with 4+ years of experience, specializing in Event-Driven Architectures, Microservices, Distributed Systems, and building secure and scalable Cloud-Native Solutions.
Project Type:
Content Automation Tool
Core Technologies:
Python, GPT-3.5, FFmpeg
(Private Instance)
Technical Overview
A weekend project that got out of hand - automates video creation by combining stock footage with AI-generated voiceovers and captions.
Workflow: User inputs topic → Python scrapes news/RSS feeds → GPT-3.5 writes script → AWS Polly generates voiceover → FFmpeg stitches clips → Uploads to YouTube.
Cool Hack: Created a "visual rhythm" system that matches clip transitions to voiceover intonation using audio waveform analysis.
Technical Challenges
Video Synchronization
- Problem: Voiceover and clips going out of sync for longer videos
- Solution: Implemented a chunk-based processing system that handles 60-second segments sequentially
API Costs
- Problem: GPT-3.5 costs adding up quickly during testing
- Solution: Created a local cache of common responses and implemented strict character limits
Lessons Learned
- • Video processing is CPU-intensive - cloud functions are essential
- • AI APIs require careful rate limiting
- • Content moderation is harder than expected
Future Ideas
- • Add custom avatar creation
- • Integrate with TikTok API
- • Local AI model deployment
