---
title: Neural Analog Engine
emoji: 🎛️
colorFrom: blue
colorTo: purple
sdk: gradio
app_file: app.py
pinned: false
---

# 🎛️ NEURAL-ANALOG ENGINE

**Turning Large Language Models into Analog Synthesizers.**

Most AI audio tools use black-box diffusion models to generate messy, artifact-heavy audio. The **Neural-Analog Engine** takes a radically different, elegant approach:
1. We prompt a 1.5B parameter LLM (`Qwen 2.5`).
2. We use PyTorch **Forward Hooks** to intercept the hidden state activations of its 15th neural layer in real-time as it generates text.
3. We pipe these raw, floating-point "thoughts" directly into a digital signal processing (DSP) oscillator.

The result is pure, artifact-free, analog-style ambient drones driven by the actual cognitive processing of an AI. 

---

### 🎵 Powered by Livadies
Created to explore the intersection of machine cognition and acoustic art.
* 🟢 [Spotify](https://open.spotify.com/artist/0j8EmbhNFjiVhIJcZHdfUD)
* 🔴 [YouTube](https://music.youtube.com/channel/UCe6BJsKd0uj1kAQcdHqyXQw)
* 🟡 [Yandex](https://music.yandex.ru/artist/21918652)

🔥 Active Project Baseline: **«RUSSIAN WINTER 26»**