A lossless-by-default codec family for audio, image, and video — built on a shared φrANS entropy core, Fibonacci lifting transforms, and a Lucas-number quality ladder. Byte-identical output across every supported architecture.
Every codec defaults to Q1 — lossless, with exact round-trip reconstruction guaranteed for all signal types. Lossy modes are opt-in, using a quality ladder derived from Lucas numbers rather than arbitrary tuning parameters.
All four codecs are blockchain-deterministic: given identical input and quality setting, the encoded byte stream is bit-for-bit identical across every supported architecture and platform. Reproducibility is a hard design requirement, not a best-effort property.
The core lossless library — φrANS entropy coding, Triveni lifting, and block classification — is shared across all three signal codecs. Audio, image, and video each add a domain-specific prediction stage on top of a common algebraic substrate.
The family is organized as a dependency stack. The core library provides the shared entropy coder, lifting transform, and block classification that every signal codec relies on. Each domain codec adds its own prediction stage and block geometry on top of that shared foundation.
Streaming-first lossless audio codec for PCM — 987-sample blocks (F(16)), M/S decorrelation, 1D Fibonacci polyphase lifting, and an 8-mode Purusha predictor. Offline and real-time streaming paths.
Spatial image codec over 21×34 golden-rectangle blocks — reversible colour transform, Phyllotaxis golden-angle prediction, Triveni 2D lifting, and Purusha rate-distortion mode selection per block.
Temporal video codec with I/P/B hierarchy — golden-angle spiral motion search, φ-interpolated sub-pixel compensation, GOP size F(8)=21, and spatial encoding delegated entirely to the image codec pipeline.
All three signal codecs share an identical quality ladder. Each tier's quantization step is a Lucas number — the same sequence that governs the φCrypt security tiers. Q1 is always lossless. Tiers increase by the same ratio that governs the rest of the framework, giving predictable quality gradations without arbitrary tuning.
The default quality. Perfect reconstruction guaranteed. Identical output across every architecture. Used for archival, blockchain payloads, and any context where round-trip exactness is required.
Minimal perceptible loss on most material. Suitable for preview generation, streaming at constrained bandwidth, and applications where slight deviation from lossless is acceptable.
Visible quality reduction; high compression gains. Audio speech targets ≤0.45× original size; image targets ≤0.28×; video targets proportional to scene complexity. The primary production lossy tiers.
Maximum compression at significant quality cost. Audio speech targets ≤0.30×; image ≤0.14×. Reserved for bandwidth-critical or storage-constrained contexts where quality is secondary.
All components are published under AGPLv3. For commercial deployment in closed-source products, a commercial patent license is available.