A universal, visual-native signal format that stores audio and other waveforms as a reversible multi-band time–frequency surface.
Status: Draft / Experimental
AuralGlyph is a new kind of signal representation that unifies audio, bioacoustics, ultrasonic/infrasonic data, sonar, vibrations, and even electromagnetic time-series into a single, coherent format.
Unlike traditional audio files, AuralGlyph represents signals as a Canonical Audio Surface (CAS) - a 3D time–frequency tensor that is:
- Reversible → can reconstruct the original signal
- Visual-native → can be displayed as images or 3D surfaces
- Multi-band → supports audible, infrasonic, and ultrasonic content
- Coordinate-agnostic → can be rendered in linear, spiral, Hilbert, cylindrical, or custom views
- Cross-domain → not restricted to human hearing or audio playback systems
AuralGlyph bridges the gap between:
- audio engineering
- bioacoustics
- scientific analysis
- signal processing
- machine learning
- digital art
In this version of the specification, a CAS instance is intended to represent one scalar-valued signal from one physical or logical sensor. Multi-sensor or multi-component setups SHOULD be modeled as collections of CAS instances, with their relationships defined at the container or session level.
Future revisions MAY introduce aggregate types or profiles that allow vector-valued signals, without changing the semantics of existing scalar CAS instances.
Existing audio formats store signals as waveforms (time-domain PCM) or psychoacoustic encodings (MP3, AAC).
Scientific formats store raw samples but not time–frequency structures.
Image formats visualize spectrograms but cannot be reversed back to audio.
AuralGlyph is the first format that is natively time–frequency based.
It provides:
- A unified representation for any sampled physical phenomenon
- Direct visualization without decoding
- Optional extended bands for non-human frequencies
- A consistent structure for ML models (spectrograms are already the lingua franca of audio ML)
- A platform for new artistic and scientific visualizations
At the heart of AuralGlyph is the CAS, a 3D tensor:
Where:
- b - band index (infra / human / ultra / custom)
- t - time frame index
- f - frequency bin index
- Each value is a complex coefficient (magnitude + phase)
This single data structure is the official representation of the signal.
Any visualization or projection (spectrogram, spiral, Hilbert, etc.) is derived from CAS without altering it.
- Infra (<20 Hz)
- Human audible (20 Hz – 20 kHz)
- Ultrasonic (20 kHz – 100+ kHz)
- Arbitrary custom bands
- Easily extended for sonar, RF, seismic, EEG/MEG, etc.
- CAS contains enough information (given parameters) to reconstruct time-domain signals
- Inversion requirements defined explicitly in the spec
- Works with STFT, CQT, wavelet transforms, or future transforms
AuralGlyph is designed to look like:
- A spectrogram
- A spiral cochleagram
- A “piano roll wheel”
- A Hilbert image
- A cylindrical sonar map
All without losing reversibility.
Long signals are stored with:
- time-indexed frames
- optional tiling
- optional multi-resolution overviews
The storage remains a clean tensor; the display decides how to render it.
AuralGlyph is not limited to audio.
It can represent:
- Whale songs
- Bat chirps
- Seismic tremors
- Biomedical signals
- Machinery vibrations
- Radio-frequency time-series (after downconversion)
- Any sampled phenomenon with a frequency structure
AuralGlyph is currently in the design and specification stage.
This repository contains:
- The evolving AuralGlyph Specification (Draft v0.1)
- Design notes and discussions
- Planned transform profiles
- Placeholder sections for future reference implementations
Nothing here is final yet - expect rapid iteration and breaking changes.
auralglyph-spec/
├─ README.md # Project overview (this file)
├─ spec/
│ ├─ auralglyph-spec-v0.1.md # Main draft specification
├─ notes/
│ ├─ design-journal.md # Rough ideas, explorations, sketches
│ └─ concepts.md # Coordinate systems, transforms, projections
├─ examples/
│ ├─ example-cas-diagrams/
│ └─ example-data/
└─ proto/ # (Future) reference encoder/decoder experiments
- Canonical Audio Surface formal definition
- Band profiles (infra/human/ultra)
- Transform requirements (STFT baseline)
- Metadata schema
- Container structure
- Linear spectrogram view
- Spiral & cylindrical projections
- Hilbert curve view
- Multi-scale / overview modes
- Python encoder (CAS ← audio)
- Python decoder (audio ← CAS)
- CAS visualizer (various projection modes)
- Music encoding
- Bioacoustic datasets
- Sonar logs
- Machine learning integration
- Scientific datasets
- Seismic data
This is an open experimental project.
Ideas, questions, and feedback are welcome.
If you want to participate in spec design, feel free to open an issue or discussion.
AuralGlyph reimagines what a “signal file format” can be:
- not just for listening,
- not just for visualization,
- not just for analysis,
but a single, unified medium that encodes the full shape of sound and other time–frequency signals -
in a form that is both human-visible and machine-reversible.
Portions of this project’s early documentation and conceptual development were created with the assistance of AI tools. These tools helped refine ideas, expand technical discussions, organize drafts, and accelerate the writing process.
The core concept, goals, and architectural direction of AuralGlyph originate with the project’s creator. AI was used as a collaborative tool - similar to a technical editor - to explore alternatives, clarify reasoning, and improve structure and consistency.
All final decisions, interpretations, and design choices were made by the project founder.
AI involvement is disclosed for transparency, and to ensure contributors understand that while AI participated in the drafting process, the creative and conceptual ownership of AuralGlyph remains firmly human-directed.