AudioIntelligence leverages the industrial-grade AVFoundation and Accelerate frameworks to provide seamless support for all professional audio codecs with zero-copy efficiency on Apple Silicon.
Through our AudioLoader orchestrator, we support any format recognized by the Apple ecosystem, including but not limited to:
| Format | Extension | Bit-Depth | Native Hardware Support |
|---|---|---|---|
| PCM (Wave) | .wav |
16, 24, 32-bit | ✅ Bit-exact passthrough |
| FLAC | .flac |
Up to 24-bit | ✅ vDSP Optimized |
| ALAC | .m4a |
Lossless | ✅ Core Audio decoder |
| MP3 | .mp3 |
Variable/Fixed | ✅ Core Audio Hardware Decoder |
| AAC | .m4a, .aac |
Lossy | ✅ Core Audio Hardware Decoder |
| AIFF | .aiff |
16/24/32-bit | ✅ Native |
Support follows whatever the host's AVFoundation/Core Audio can decode. Container formats that the Apple stack does not natively decode (e.g. Ogg Vorbis/Opus) are not supported.
Unlike general-purpose research libraries, AudioIntelligence is engineered for Industrial Forensic Fidelity:
- 32-bit Floating Point Foundation: All internal engines process data in
Float32orFloat64to prevent rounding errors. - Upsampling audit: the
ForensicEnginedetects "fake hi-res" by comparing the declared container bit depth against the measured effective bit depth (not by entropy). - SQAM Parity: Validated against the EBU Sound Quality Assessment Material.
- AES17 Compliance: Dynamic range and frequency response metrics adhere to AES17-2020 standards.
Our loading pipeline is not just about reading bytes; it's about preparation:
- Automatic Down-mixing: Robust stereo-to-mono or multi-channel handling via
AVAudioConverter. - Intelligent Resampling: High-quality polyphase resampling to the target engine rate (default 22050Hz or 44100Hz).
- IntelligenceCache: Once a file is fingerprinted (SHA-256), its spectral representations are cached in a 4GB hybrid buffer for instantaneous subsequent access.