How Ecasound Processes Multi-Track CAF Files

This article explains how the Ecasound command-line audio utility handles multi-track Core Audio Format (CAF) files. It covers the underlying libraries responsible for decoding the CAF container, how multi-channel streams are split into processing chains, and the specific routing options available for manipulating multi-track audio effectively.

Decoding CAF Files via libsndfile

Ecasound does not parse the Core Audio Format specification directly; instead, it relies on libsndfile as its underlying audio file I/O engine. Because modern versions of libsndfile support Apple’s CAF specification, Ecasound automatically recognizes .caf files upon input.

When Ecasound opens a CAF file, it queries libsndfile to detect the container's structural parameters, such as sample rate, bit depth, and total channel count. This enables Ecasound to overcome the 4 GB file size limitation common to standard 32-bit RIFF/WAV headers, making CAF ideal for recording and processing long, high-resolution, multi-track sessions.

Chain Architecture and Channel Allocation

Ecasound manages audio streams using an architecture based on "chains." A multi-track CAF file contains interleaved audio channels within a single container. To process these channels independently, Ecasound maps the file’s internal tracks to distinct processing chains using the -a option.

By default, an input file assigned to a single chain reads all channels simultaneously. However, when dealing with multi-track CAF files, you typically route individual channels or sub-mixes into separate chains using channel-routing operators:

• -chcopy:src,dst: Copies an audio channel from an input slot to an internal processing slot.
• -chorder:ch1,ch2,...: Reorders or selects specific channels from the input stream.
• -chmove:src,dst: Moves a channel, clearing the source slot.

Practical Multi-Track Workflow

To work with a multi-track CAF file, you instantiate the file as an input across multiple chains, then configure channel filtering on each chain.

For example, processing a 4-channel CAF file by splitting tracks 1 and 2 into one stereo chain and tracks 3 and 4 into another is structured as follows:

ecasound \
  -a:1 -i:multitrack.caf -chorder:1,2 -o:stereo_mix1.wav \
  -a:2 -i:multitrack.caf -chorder:3,4 -o:stereo_mix2.wav

In this command:

1. -a:1 creates the first chain, defines multitrack.caf as the input, and selects channels 1 and 2 with -chorder:1,2.
2. -a:2 creates the second chain, accesses the same file simultaneously, and isolates channels 3 and 4 with -chorder:3,4.
3. Ecasound reads the interleaved CAF stream, and its internal routing engine dispatches the respective audio samples to the assigned chain buffers before rendering the output.

Real-Time Handling and Buffering

When running multi-track CAF processing in real-time, buffer management is controlled using the -b:buffersize and -z:feature flags (such as double-buffering or real-time priority). Because multi-track audio multiplies the data throughput per processing cycle, Ecasound reads interleaved multi-channel blocks into memory, de-interleaves them for internal chain effects or dynamic DSP plugins (such as LADSPA), and re-interleaves them if the output format requires it.