Projects — Real AI Systems and Video Pipelines
This page shows practical systems I build using AI.
Not demos.
These are real pipelines designed for: - reliability - repeatability - production use
1. WorkplaceSafetyAI — AI Safety Video Pipeline
Short-form videos based on real workplace incidents.
Goal: - explain risks visually - make safety training engaging - scale content production
Pipeline
Prompt → Image → Video → Voice → Lip Sync → Final Video
Tools: - ComfyUI (image generation) - Wan / LTX (video) - VoxCPM / TTS (voice) - ffmpeg (rendering)
Features
- consistent characters
- realistic scenes
- short-form optimized (YouTube Shorts, TikTok)
- repeatable workflow
Example Scenario
- construction worker on unstable scaffold
- unsafe forklift operation
- lack of PPE
Output
- vertical videos (1080x1920)
- 10–30 seconds clips
- narration + visuals
2. AI Quirks & Comedy
AI-generated videos highlighting unexpected or funny model behavior.
Goal: - show AI limitations - create viral content - build audience
Example
Prompt:
man brings flowers to his wife
Result:
AI makes him run into a wall
Value
- engaging content
- high share potential
- demonstrates AI behavior
3. Character Consistency System
Problem: AI characters change between scenes.
Solution: - fixed prompts - reference images - LoRA models - controlled generation
Result
- same face across scenes
- consistent identity
- usable for storytelling
4. AI Infrastructure (GPU + Docker)
I build local AI systems using:
- GPU servers
- Docker stacks
- reverse proxy
- storage systems
Example Setup
Proxmox
├── GPU Node (ComfyUI)
├── AI Services (OpenClaw, Ollama)
├── Storage (NFS)
└── Web (MkDocs, WordPress)
What This Enables
- local model execution
- no dependency on external APIs
- predictable performance
- data control
5. Automation Systems
Goal: reduce manual work.
Examples
- batch video generation
- automatic rendering
- scheduled publishing
- workflow orchestration
Tools
- bash scripts
- Python
- n8n
- API integrations
6. End-to-End AI Video System
Full pipeline:
Prompt → Character → Image → Video → Voice → Lip Sync → Export
This allows: - content at scale - consistent style - reusable assets
7. What I Focus On
I don’t build random experiments.
I build systems that: - can run daily - can be scaled - can be maintained
8. Who This Is For
- content creators
- startups
- marketing teams
- AI teams building pipelines
9. What I Can Build For You
- AI video pipelines
- GPU infrastructure
- automation workflows
- custom AI systems
10. Next Step
If you want to build something similar: