AVELUNE
historical · non-normative

Avelune Audio Candidate 0.0

Historical Avelune audio codec experiments, retained for provenance and superseded by Draft Generation 1.

View specification source
Specification navigation

Current draft

  1. Avelune specification drafts
  2. Avelune Draft Generation 1 Baseline profile
  3. ALV1 video bitstream — Draft Generation 1
  4. Avelune V1 — common integer and entropy coding
  5. Avelune container — Draft Generation 1 (header generation 1, minor 0)
  6. ALA1 audio bitstream — Draft Generation 1
  7. Avelune Draft Generation 1 conformance requirements
  8. Specification changelog

History

  1. Avelune Audio Candidate 0.0
  2. Avelune Container Candidate 0.0
  3. Avelune Video Candidate 0.0
  4. Candidate conformance requirements
On this page
  1. Intended architecture
  2. Experiment 0: lossless predictive PCM
Warning

HISTORICAL / NON-NORMATIVE: superseded by the Draft Generation 1 001-* specification.

Status: EXPERIMENTAL / NOT FROZEN

Intended architecture

The main research direction is a general-purpose lapped transform codec with explicit band energies and optional short-term/pitch whitening before the common transform path. It is not frozen.

Initial target configuration is 48 kHz mono/stereo with nominal 20 ms framing and shorter transform partitions for transients.

Experiment 0: lossless predictive PCM

The executable bring-up format deliberately starts with a simple custom codec rather than embedding an existing codec or treating PCM as the final design.

Input samples are signed 16-bit integers. For each channel independently:

  1. the first sample is predicted as zero;

  2. subsequent samples are predicted by the immediately previous reconstructed sample for the same channel;

  3. the signed residual is mapped by ZigZag coding;

  4. the unsigned value is encoded as bounded LEB128.

Predictor state MUST reset at every epoch boundary.

This experiment is useful for packetization, synchronization, truncation handling and exact round-trip tests. It is not the intended lossy audio architecture.