Activity Feed

AI & ML interests

None defined yet.

Recent Activity

Bils 
posted an update 3 days ago
view post
Post
2607
Avatars are everywhere, but here is the reality behind full-system marketing automation. 🚀
Many see "Madame AI" simply as an AI news presenter. She is far deeper than that. Madame AI is a Real-time Agentic AI Assistant we developed to orchestrate entire workflows for marketing and professional media. She manages UGC (User-Generated Content), understands marketing system automation intuitively, and handles complex media tasks.
We have solved the character consistency and high production cost bottlenecks that traditionally required immense training and time. By precisely orchestrating every computational step behind videos and branded designs, we have fully automated the pipeline and significantly reduced costs.
This capability is built on our extensive experience managing large-scale automation projects with complex requirement documentation (PRD).
Grabclip is our public portal and the practical result of that journey. It is the interface where "Madame AI" acts as the intelligent engine.
We have spent three years building this pipeline with a clear goal: a 100% local, end-to-end solution that operates despite external restrictions.
See the live example on YouTube (our fast-paced AI news podcast with Madame AI) and try the automation portal yourself👇
📺 The Playlist: https://www.youtube.com/playlist?list=PLwEbW4bdYBSCVSziFfJYq4zXop_cyHquO
🌐 Our Portal (Grabclip) — The first practical step in our pipeline: https://grabclip.bilsimaging.com/
hashtag#AgenticAI hashtag#VirtualInfluencer hashtag#FutureOfWork hashtag#GenerativeAI hashtag#TunisiaTech hashtag#MarketingAutomation hashtag#100PercentLocal hashtag#OSMedia hashtag#Grabclip hashtag#RealTimeAssistant hashtag#UGC hashtag#ProfessionalMedia hashtag#TunisiaAI
Severian 
posted an update 14 days ago
view post
Post
4400
I’ve been working on a new mathematical approach to real-time video compositing and background removal, and I wanted to share a live demo.

Traditionally, real-time keyers either use 3D color-space bounding boxes (which struggle with semi-transparent hair and motion blur) or heavy Machine Learning models (which require massive GPU compute and often suffer from temporal "jitter" on the edges).

I wanted to see if I could solve this using purely deterministic math so it could run client-side in a standard browser.

The engine uses a custom mathematical framework I call CMT SRL SEFA. Instead of looking at raw color values or guessing semantics like an AI, it treats the video feed as complex-encoded sequences. It uses harmonic frequencies to map phase geometry and applies a "Stability Cost Function" to find the global minimum stability. In short: it isolates the foreground from the background by measuring signal complexity and structural contradictions.

Give it a try using your own messy plates and such. As I am not a VFX artist, I am curious to hear thoughts and what should be improved upon and made better

https://huggingface.co/proxy/severian-cmt-sefa-realtime-vfx-keyer.hf.space/
  • 2 replies
·
Bils 
posted an update 3 months ago
view post
Post
322
We just published a workflow that automates trend-style celebrity selfie videos — from image generation to cinematic transitions.

This template is a mini creative factory:
→ Generate realistic “celebrity selfie” images
→ Produce clean, cinematic transitions ready for Shorts/Reels
→ Clear structure, easy to customize for your brand

📌 Template link:
https://n8n.io/workflows/12119-create-celebrity-selfie-images-and-transition-videos-with-gpt-4-seeddream-and-kling/