š¤ Automating Creativity: 1 Photo + Code = Full Promo Video
This video was generated 100% programmatically using Remotion and Qwen3-TTS. No manual editing, no voice actors.
I built a pipeline that takes a static product photo, clones the brand's voice for a dynamic voiceover, and renders a perfectly synced 1280x720 video in under 60 seconds.
š ļø Stack: TypeScript, Remotion, Python, PyTorch (ROCm), Cloudflare R2.
#ProgrammaticVideo #AIEngineering #Automation #Remotion #TypeScript
In the auto parts industry, inventory is massive (thousands of SKUs) but visually static. Sellers have photos, but social platforms (TikTok, Reels, YouTube Shorts) demand engaging video content.
Manual editing was impossible at scale:
Cost: ~$50 per video for a human editor.
Time: Hours of turnaround.
Voiceover: Expensive and slow to coordinate.
We needed a way to turn 1 photo + data into a professional promo video instantly.
š” The Solution
I engineered a Programmatic Content Factory that automates the entire creative process using code.
1. AI Voice Cloning (Qwen3-TTS)
Instead of generic robot voices, I integrated Qwen3-TTS to clone our brand's specific voice signature.
Input: Text script generated from product data.
Output: Human-like WAV file with proper intonation.