100% On-Device • Zero Network Calls • iOS 17+

Frontier-Level Audio.
In Your Pocket.

Nine interactive labs for real-time DSP, spectral analysis, spatial audio, neural enhancement, and professional measurement — all running on a lock-free C++ engine with zero network calls.

Download on the App Store Explore Labs
Free with Biquadia Pro $4.99/mo • $79.99 Lifetime
See It in Action

Nine Labs.
One Engine.
Zero Compromise.

Watch Biquadia's real-time C++ DSP engine process live audio — from spectral analysis and neural voice isolation to spatial beamforming — all running on-device with zero network calls.

  • Lock-free render thread — never drops a frame
  • Apple Neural Engine acceleration at 21ms latency
  • Professional measurement: THD+N, SNR, calibrated SPL
Download Free

How It Works

From microphone to measurement — a complete audio engineering pipeline on your iPhone.

Input Mic • File • USB-C
→
DSP Graph C++ Lock-Free Engine
→
Analysis & Export Visualize • Measure • Compare

Engineered for Precision

A stunning UI packed with a true C++ DSP engine, spatial computing, and Apple Neural Engine acceleration.

Zero-Latency DSP

Real-time FIR/IIR/Biquad/Butterworth/Chebyshev/Linkwitz-Riley filtering, compressor/limiter/gate/expander dynamics, and delay-family effects — all on a lock-free C++ render thread that never drops a frame.

Neural Enhancement

Voice isolation via Apple Neural Engine with chunked OLA streaming. Target-speaker extraction using 128-d speaker embeddings and causal LRU layers — 21ms latency at 48kHz.

Advanced Array Processing

True covariance-based MVDR beamforming with EMA noise estimation, GCC-PHAT DOA, and N-mic SRP-PHAT — up to 32 channels over USB-C.

Ambisonics & Spatial

First-Order Ambisonics (ACN WXYZ / FuMa) with energy-preserving rotation, HRTF binaural rendering, AirPods head tracking, and virtual-speaker decode at ±30°/±90°.

9 Interactive Labs

Each lab is a complete surface: parameters, visualization, and measurement — designed for learning and professional use.

DSP Laboratory interface showing oscilloscope and FFT controls

DSP Laboratory

Live oscilloscope, configurable FFT spectrum, FIR/IIR filter design with magnitude/phase/group-delay/pole-zero visualization.

FFT Biquad Dynamics Delay
Analysis Laboratory with STFT spectrogram and pitch tracking

Analysis Laboratory

STFT/Mel/Log spectrograms, autocorrelation and YIN pitch tracking, LPC formants, MFCC delta/delta-delta, spectral descriptors.

STFT MFCC Pitch LPC
Spatial Lab with HRTF binaural room and FOA rotation controls

Spatial Lab

HRTF binaural rendering, early reflections + FDN reverb, room/source/listener controls, AirPods head tracking, FOA encoding and rotation.

HRTF FOA Binaural
Enhance Lab showing neural voice isolation and SPP curves

Enhance Lab

Wiener, MMSE, Ephraim-Malah with SPP, ANE Core ML voice isolator, target-speaker extraction with 128-d d-vector FiLM conditioning.

Wiener MMSE ANE TSE
Codec Lab with difference spectrogram and A/B blind comparison

Codec Lab

WAV/FLAC/AAC/Opus/MP3 decode, AAC 48–256kbps + Opus encode, bitrate visualization, compression ratio, difference spectrogram, A/B blind.

AAC Opus A/B Blind
System Monitor showing compute cost RTF and I/O channel mixer

Beam & Array Processing

GCC-PHAT DOA, delay-and-sum, true MVDR beamforming (covariance Rnn, EMA α=0.05), N-mic SRP-PHAT up to 32ch USB-C, dual-binaural HATS.

MVDR GCC-PHAT USB-C
System Monitor with RTF compute cost, signal-path latency, and thermal state

System Monitor

Mach absolute-time whole-graph RTF, signal-path latency ledger, CPU/RAM/thermal/thread monitoring, per-channel peak meters (32ch), hardware I/O card.

RTF Latency Thermal
ML Sandbox showing Core ML inference timeline and speaker embeddings

ML Sandbox

Core ML FP16 inference, chunked/OLA streaming, 128-d L2-normalized speaker embeddings, 2-speaker heuristic diarization.

Core ML FP16 Embeddings
Measurement surface showing THD, SNR, and frequency response analysis

Measurement

THD/THD+N/SNR/DR/frequency response, impulse and log sweep generation, transfer function estimation, SPL and reference calibration.

THD SNR Sweep SPL

Built For

Whether you're shipping a product, publishing a paper, or learning the fundamentals.

Audio Engineer

  • Real-time FIR/IIR filter design with pole-zero plots
  • True MVDR beamforming over USB-C multi-mic arrays
  • THD/THD+N/SNR/DR measurement with log sweep
  • Signal-path latency ledger and compute cost RTF

Researcher

  • MFCC/delta/delta-delta, LPC formants, cepstrum analysis
  • 128-d speaker embeddings with L2 normalization
  • FOA encoding with energy-preserving rotation
  • SI-SDR / SegSNR offline evaluation metrics

Student

  • Interactive oscilloscope and FFT with adjustable windows
  • Magnitude/phase/impulse/step response visualization
  • Codec A/B blind comparison: hear the difference
  • Learn Catalog with lab browser and guided exploration

Technical Specifications

Under the hood of our highly optimized C++ and CoreAudio stack.

DSP Engine Lock-free real-time C++ kernel
Audio I/O CoreAudio / AVAudioEngine (USB-C interface support up to 32ch)
Neural Inference Core ML / Apple Neural Engine / FP16
Beamforming True MVDR (covariance Rnn, EMA α=0.05, diag 1e-4) + Delay-and-Sum
Spatial FOA (ACN WXYZ / FuMa) + HRTF binaural + AirPods head tracking
Codecs WAV / FLAC / AAC (48–256kbps) / Opus (CAF) / MP3
Measurement THD / THD+N / SNR / DR / Frequency Response / SPL Cal
Math Accelerate framework (vDSP)
Privacy 100% on-device — zero network calls, zero telemetry
Hardware iOS 17.0+ • iPhone / iPad

Frequently Asked Questions

Real answers to real audio engineering questions.

Error -10875 (kAudioUnitErr_CannotDoInCurrentContext) occurs when iOS revokes the audio route — typically when backgrounding the app while the engine is running. Biquadia handles this automatically: it saves your run state, stops the engine on background, and restarts when you return. If you see this after a USB-C device change, tap the Refresh I/O button in the Session sheet.
No. Biquadia is 100% on-device with zero network calls. All DSP, neural inference, and analysis run locally on your device's CPU and Apple Neural Engine. Your audio never leaves your phone. The microphone permission is used solely for live input — there is no telemetry, analytics, or cloud processing.
Biquadia Pro is managed entirely through Apple's App Store. To restore: open Settings → tap the Pro badge or subscription row → tap "Restore Purchases." Make sure you're signed in with the same Apple ID used for the original purchase. If the restore doesn't work, check Settings → Apple ID → Subscriptions on your device.
Connect a class-compliant USB-C audio interface (e.g., Behringer UMC1820). Open Biquadia → Session sheet → select your interface under Input Device. The I/O channel mixer shows all available channels with per-channel peak meters. For beamforming, go to the Beam lab and configure your array geometry (linear or circular). Note: some interfaces require a manual "Refresh I/O" tap after connecting due to USB-C power negotiation quirks.
Biquadia does not block screenshots. If a screenshot appears black, it's because iOS captures the screen at a moment when the Metal or GPU-rendered visualization hasn't composited. Try pausing the engine first (tap STOP), then take your screenshot. Alternatively, use iOS Screen Recording for a video capture of live analysis.
Biquadia requires iOS 17.0 or later. It runs on any iPhone or iPad with an A12 Bionic chip or newer. Neural Enhancement features (voice isolation, target-speaker extraction) benefit from devices with a dedicated Neural Engine — iPhone 15 Pro and later deliver the best ANE performance.