Audix #

The intelligent, high-fidelity audio engine for Android. Powered by AI, tuned for perfection.

Audix is a premium audio enhancement suite that automatically adapts your device's sound signature based on real-time genre analysis. It bridges the gap between professional-grade audio processing and seamless, zero-config user experience.

Interface Jetpack Compose
Intelligence Gemini 2.5 Flash Lite
Engine Android SDK 34+
Priority Low-Latency DSP

Getting Started #

Audix is designed to be invisible. Once set up, it lives in the background, enhancing every note you play.

Installation: Audix requires Notification Access to track media metadata and Battery Optimization Exemption to prevent service termination during long listening sessions.
  1. Grant Permissions: Enable Notification Listener and Background Activity in the welcome screen.
  2. Sync Engine: Audix will automatically attach to Session 0 (Global) to capture all system audio.
  3. Listen: Simply play music from any supported app; the AI will take over within milliseconds.

Supported Players #

Audix uses advanced IPC (Inter-Process Communication) to extract metadata from the following players:

Spotify Native Support
YouTube Music Native Support

AutoEQ Engine #

The AutoEQ engine is the core differentiator of Audix. It moves beyond static presets into adaptive, context-aware sonic profiles.

Adaptive Gain Compensation

Standard Android EQs often reduce overall volume to prevent clipping. Audix implements a +3.0dB pre-gain offset with soft-limiting to maintain high loudness while applying rich curves.

Smart Fade: When genres transition (e.g., from Jazz to Metal), Audix smoothly interpolates EQ bands over 500ms to avoid abrupt sonic shifts.

Spatial Audio #

Audix's spatial engine uses a 4-layer psychoacoustic approach to create a 3D soundstage from a standard stereo source.

  • Layer A: Pinna Notch: Simulates the frequency response of the human ear to improve 3D positioning.
  • Layer B: Stereo Widening: Expands the virtual field beyond the physical driver constraints.
  • Layer C: Environmental Depth: Adds sophisticated reverb models (Room, Concert Hall, Stadium).
  • Layer D: Bass-Aware Dynamics: Preserves low-end clarity even at high virtualization levels.
Headphones Required: Spatial algorithms are phase-sensitive and require headphones to work effectively. Audix auto-disables this feature on internal speakers.

Custom Tuning #

For purists, Audix offers high-precision manual controls that override the AI engine for the ultimate personalized signature.

Bass Low-Shelf ≤ 250Hz
Vocals Bell 800 - 3kHz
Treble High-Shelf ≥ 8kHz

System Architecture #

Audix is built on a modern, decoupled architecture designed for high availability and low latency.

The Layered Approach

  • Orchestrator: AudioEngineServiceLocal handles the foreground lifecycle and global audio session binding.
  • Observer: SongDetectionService utilizes Android's NotificationListenerService to track and react to media changes across the system.
  • Engine: A high-performance wrapper around the AudioEffect API, managing Equalizer and Virtualizer instances with priority 1000.
  • Data Layer: Room-backed persistence for genre caching, ensuring millisecond response times even when offline.

DSP Pipeline #

Audix applies a rigorous processing chain to every audio buffer passing through the system.

  1. Pre-Analysis: Genre is identified via AI or Local Cache.
  2. Gain Preparation: Signal is padded by 3dB to provide headroom for EQ boosting.
  3. Psychoacoustic Profiling: Pinna notch filters are applied if Spatial Audio is active.
  4. Curve Application: 10-band hardware EQ is programmed with the target signature.
  5. Environment Sim: Stereo widening and environmental reverb are applied at the tail.
  6. Limiting: Soft-clipper ensures no digital distortion occurs at peak amplitude.

AI Intelligence #

Audix leverages a hybrid intelligence model to balance speed and accuracy, primarily powered by the next-generation Gemini 2.5 Flash Lite model.

Gemini 2.5 Flash Lite Integration

When a new song is detected, Audix sends an encrypted metadata packet to our classification model. Using Gemini 2.5 Flash Lite's industry-leading latency and massive context window, the model identifies not just the genre, but the "vibe" and "sonic density" to pick the perfect EQ profile in under 200ms.

API Key Management

Audix provides a seamless experience regarding AI connectivity. You can choose between the built-in engine or your own dedicated pipeline.

  • Built-in Engine (Default): Audix comes with a shared developer key. It's zero-config but may be subject to global rate limits during peak usage.
  • Dedicated Key (Recommended): By providing your own Google AI Studio API key, you unlock private quotas, higher rate limits, and zero latency overhead from shared queues.
How to configure: Navigate to Settings > Advanced > API Key in the app to paste your own Gemini key.

The Local Fallback

Metadata for over 50,000 common tracks is stored in an optimized local Room database. If classification fails, connectivity is lost, or the built-in key hits a rate limit, Audix falls back to local knowledge in < 5ms.

Troubleshooting #

Technical solutions for common edge cases.

No effect in specific apps (e.g., Poweramp)

Some premium players use their own audio engine and bypass the system effect chain. Ensure "MusicFX" or "System Audio Effects" is enabled in your player's settings.

OEM Conflicts (Xiaomi Sound, Oreality audio, Dolby Atmos etc.)

Audix uses Priority 1000 to override system effects. If you bear distortion, try disabling the OEM audio suite (Xiaomi Sound, Oreality audio, Dolby Atmos etc.) to prevent filter stacking.

Privacy & Security #

Audix is designed with Privacy-by-Default. We don't want your data; we only want to improve your sound.

  • Metadata Only: We only process Track Title and Artist. No private notifications are ever read.
  • No Accounts: Audix works 100% anonymously. No email or login required.
  • Local Logic: All DSP processing happens locally on your device's audio hardware.
  • Open Source: Our code is AGPL-3.0 licensed, allowing full community auditing.

Contributing #

Help us make world-class audio accessible to everyone. We welcome contributions to the Android app, the AI modeling, and our documentation.

Visit our GitHub Repository to get started.