Independent developer · Italy
Hi 👋 I build open-source AI tools that run on your own machine — no cloud, no API keys, no subscription.
metiu.xyz — projects, contact form, and a place to pitch me an idea.
Vortelio — run every AI model on your own PC
The open-source alternative to Ollama + ComfyUI + Whisper UI + CrewAI cloud, in one binary.
Chat with an LLM, generate images, transcribe audio, synthesise speech, make video and 3D — all locally. No cloud, no API keys, no subscriptions.
- OpenAI-compatible and Ollama-compatible API, so existing code just works
- Go server + Python backends (diffusers, faster-whisper, kokoro)
- Single binary, auto-detects CUDA / ROCm / Metal / CPU
go python llama.cpp stable-diffusion whisper crewai openai-compatible self-hosted
TubeAssistant — a YouTube channel that runs itself
An AI agent that does the whole job: picks the topic, writes the script, records the voice, edits the video, designs the thumbnail, uploads it. Every day, on its own. You control it from Telegram.
- 15 AI providers supported (OpenRouter, OpenAI, Gemini, Groq, DeepSeek, local Ollama…)
- Runs entirely on free tiers — $0/month
- Windows, Linux, macOS
python ai-agent youtube-automation llm tts openrouter telegram-bot
vedit — tell an AI to edit your video
A normal video editor — timeline, clips, cuts, transitions — except an AI can drive it for you. You describe the edit in plain words and watch the clips move, then take the mouse and fix anything you want.
- 77 tools an agent can call — the same operations as the buttons, nothing agent-only
- Works from Claude Code, Cursor or Codex via MCP: one line to install
- Deterministic ffmpeg render, runs 100 % locally
python react ffmpeg mcp ai-agent video-editing nle
| Project | What it does |
|---|---|
| vortelio-python | The Python SDK for Vortelio — chat, images, speech, video and 3D on your own machine, one line each. Zero dependencies, OpenAI/Ollama compatible. |
| rvm-web | Cut the background out of a video without a green screen, on your own PC. Real alpha transparency, plus AI upscaling and image/video editing. |
| lithophane-studio | Turn a photo into a 3D print that glows in backlight. Browser-only, nothing uploaded, exports print-ready STL/OBJ. |
| acf-agent-format | A compact replacement for JSON in an LLM's context window. Same data, 46 % fewer tokens, lossless roundtrip. |
| llm-tier-router | A ~1 ms classifier that sends easy prompts to the cheap model and hard ones to the expensive one. 97.6 % accuracy, 12k-example dataset included. |
| Languages | Go · Python · JavaScript / TypeScript |
| AI / ML | llama.cpp · diffusers · faster-whisper · sentence-transformers · CrewAI · OpenRouter |
| Infra | FFmpeg · MCP · Telegram Bot API · YouTube Data API v3 |
| APIs | OpenAI-compatible · Ollama-compatible |
Everything you need to reach me is on metiu.xyz.
- Got an idea? There is a Pitch me something form on the site — describe what you have in mind and it lands straight in my inbox. Half-formed ideas are welcome; that is what the form is for.
- Want to work together, or hire me? Same place, contact section.
- Found a bug, or want a feature? Open an issue on the repo itself — that is faster than any form.
I read everything. I am one person, so a reply may take a few days.