theallelectricsmartgrid

Sampler-Looper Recording

The sampler-looper system gives each track a way to capture audio, turn the capture into a loop-aligned sample, and make that sample available through the normal sample-source path.

Each track has three related pieces:

a recording buffer,
a sample source,
a sample directory.

The recording buffer is fixed per voice and lives next to the component that will play it back. A track also has a configurable recording source. That source selects which trio/lane output should feed the track’s own recording buffer while recording. The sample source plays back from an AudioBufferBank. The sample directory is the persistent directory backing that bank. This complexity mirrors the Octatrack’s flex buffers, where recording buffers and playback assignments are related but not the same object.

Recording Sources

Recording is routed through RecordingConfig. A track’s source selection chooses which trio’s same-lane voice should provide audio. For example, lane 2 on a track configured for Water records Water lane 2’s output into that track’s own recording buffer. A voice can select one of the internal trio sources:

Water
Earth
Fire
None

Internal recording uses RecordingBufferWriter on the selected source voice to fan out source samples to every active destination RecordingBuffer. External sampling is a planned extension of the same model: an external source writer should feed selected destination buffers, which then persist into sample directories and reload sample banks after saved files appear on disk.

RecordingManager coordinates recording by voice or by trio. It resolves each voice’s selected source writer, starts or stops the voice’s own recording buffer, and registers or deregisters that buffer with the source writer.

Recording Buffers

RecordingBuffer is a mono in-memory capture buffer. It has a fixed capacity and an explicit state machine:

Idle
Recording
Done
Error

On start, a buffer in Idle clears previous audio and stores the current unwound master-loop position from TheoryOfTime. While recording, the selected source voice’s RecordingBufferWriter pushes one sample into every registered destination buffer each audio tick. On stop, the destination buffer stores the stop position and computes how the captured span relates to the available TheoryOfTime loops.

This start/stop phase information is what lets a recording become a loop rather than a raw slice of audio. If the recording span does not map cleanly into the supported loop range, the buffer enters Error instead of producing a file.

Loop-Aligned Persistence

When a recording is persisted, RecordingBuffer::WriteToFile() writes a mono WAV through MultichannelWavWriter.

The write is phase-aware:

If recording starts and stops within the same master-loop turn, the file is padded before and after the captured audio so the result fills the loop.
If recording wraps across the master-loop boundary, the wrapped tail is written first, silence fills the gap, and the beginning of the capture is written last.
If the chosen loop requires multiple repeats, the same loop-aligned material is written repeatedly.

The result is a sample that fits exactly one full loop of the selected time relationship, so the sample source can play it back as part of the same TheoryOfTime structure.

Sample Directories

Sample directories are relative to the configured sample root. A track with an existing sample directory writes its recording into that directory. A track without an existing directory can create a UUID-named directory under the sample root and use that directory as its sample bank.

Recorded files use the name form _recording_NNNNN.wav. The number is based on the count of regular files already in the target directory, so each recording is appended as the next sample in that directory.

Multiple tracks may share one sample directory. After recording, the system reloads known sample directories instead of trying to identify every track that points at the changed directory. Reloading is inexpensive, and the simpler rule keeps shared-directory behavior predictable.

AudioBufferBank stores each loaded sample with its source file name. Directory reloads can reuse buffers whose file names are still present, preserve the bank position, and add newly recorded files.

Sample Playback

The sample source reads from the track’s AudioBufferBank. SampleBankPosition selects or blends between files in the directory, while SampleLoopIndex, SampleStart, SampleLength, and SampleReadSpeed control how the selected sample is read through the grain engine.

The looper side and the sample source both use TheoryOfTime loop information. Recordings use the master loop to define their captured span, and sample playback uses selected Theory of Time loops for synchronized playback phase.

This site is open source. Improve this page.