“Buffer definition”; abstraction of audio buffers for Lisp sound synthesis systems.

Basically, this simplifies buffer management in cl-collider, making them easier to use.

Note: Previously Bdef for SuperCollider was hosted at this URL. That repository has since moved here.

Bdef is available from Quicklisp, so you can simply Quickload it:

(ql:quickload :bdef)

It also has multiple sub-systems for integration with various other libraries. So if you use cl-patterns, cl-collider, or Incudine you may want to load bdef/cl-patterns and/or the other sub-systems:

;; integration with cl-patterns and cl-collider:
(ql:quickload '(bdef/cl-patterns bdef/cl-collider))

To load a buffer from a file, provide the path to the bdef function:

(bdef "~/path/to/your/buffer.mp3")

;; you can also give it an alias:
(bdef :foo "/path/to/buffer.wav")

;; the buffer can then be referred to with it:
(bdef :foo)

bdef returns a bdef object which can be provided in place of a buffer for any functions that expect it, as long as you’ve quickloaded the proper bdef sub-system for the library those functions are from. For example:

;; cl-collider:
(synth :playbuf :buffer (bdef :foo)) ; will work if you've loaded the bdef/cl-collider system.

;; cl-patterns:
(play (event :instrument :playbuf :buffer :foo)) ; will work if you've loaded the bdef/cl-patterns system.
;; Notice that you can just specify the bdef name as a symbol without even having to use the bdef function! The same is true in patterns too.

Metadata can be associated with bdefs for various uses:

(setf (bdef-metadata :foo :bpm) 120) ; set the "bpm" metadatum to 120

Some metadata is automatically generated; for example, the tempo may be automatically detected from the filename or id3 tags.

bdef includes functionality for defining buffer regions called splits. This is useful, for instance, to divide a drum loop up by each hit, or to divide up a source track into sections based on onsets or beats.

;; define three consecutive regions, one second long each:
(make-splits (list (list 0 1) (list 1 2) (list 2 3)) :unit :seconds)

;; auto-generate splits from sound onsets in a file using the aubio library:
(splits-from-aubio-onsets "/path/to/file.wav")

Aubio is an external library of audio analysis functions. If installed, splits can be automatically generated from its analyses.

bdef can also generate splits from other formats as well:

• OP-1 drumsets (splits-from-op-1-drumset, but automatically parsed from any valid aif or aiff file)
• Audacity labels (splits-from-audacity-labels)
• Renoise regions (planned for future implementation)

See the following section for detail on more features of the bdef library.

Compare:

(defparameter *buf* (cl-collider:buffer-read "/buffer.wav"))

(defparameter *buf* (cl-collider:buffer-read "/buffer.wav")) ; the same variable, and same file!

(length (remove nil (slot-value *s* 'cl-collider::buffers))) ; => 2 -- duplicate buffers!

versus:

(bdef :buf "/buffer.wav") ; here we give it the name :buf

(bdef :foo "/buffer.wav") ; same file, different "name"...

(length (remove nil (slot-value *s* 'cl-collider::buffers))) ; => 1 -- no duplicate buffers :D

…To force a file to be reloaded, simply call bdef-free on it, then call bdef again.

(bdef :bar "/blah.mp3") ; works!

It does this by storing them in a temporary directory (/tmp/bdef/ by default on linux and mac).

(bdef "/my-file.ogg")

(bdef "~/cool-sound.wav") ; will find a cool sound in your home directory

For consistency. To load as mono, supply 1 for bdef’s num-channels keyword argument.

(bdef "~/wilhelm.wav" :wavetable t) ; load the Wilhelm scream as a wavetable

Either as wavetables, or standard.

(cl-collider:bufnum (bdef :sound)) ; returns the buffer number.

(cl-collider:synth :playbuf :bufnum (bdef :sound)) ; plays the buffer.

Load the bdef/cl-collider system to enable this.

(cl-patterns:play (bdef :sound)) ; plays the buffer using the *cl-collider-buffer-preview-synth* set in cl-patterns.

(cl-patterns:play (cl-patterns:event :instrument :playbuf :bufnum (bdef :sound))) ; automatically converts bdef to the buffer number.

Load the bdef/cl-patterns system to enable this.

SuperCollider/cl-collider is the primary backend tested against, however Incudine is also supported for most functionality.

Enable the cl-collider backend, for example, like so:

(ql:quickload :bdef/cl-collider)

(setf (bdef-metadata (bdef :snd) :bpm) 99) ; set :snd's tempo to 99 BPM.

(bdef-metadata (bdef :snd) :bpm) ; get the stored bpm value.

;; load a file with its bpm in its filename:
(bdef :my-file "~/my-file-128bpm.wav")

;; the bpm is automatically stored as metadata:
(bdef-metadata :my-file :bpm) ; => 128

You can also add your own auto-metadata keys with the define-bdef-auto-metadata macro or set-bdef-auto-metadata function, or remove them with the remove-bdef-auto-metadata function.

Additional metadata is loaded asynchronously in background threads using futures from the eager-future2 library. If a requested metadatum is still being generated, bdef-metadata will block until the result is available.

(setf (bdef-metadata :foo :bpm) 142) ; sets the "tempo" metadata key instead to its beats per minute value

;; tempo is stored as beats per second:
(bdef-metadata :foo :tempo) ; => 71/30 (142 beats per minute in beats per second)

;; beats per minute is still available, dynamically calculated from the tempo key:
(bdef-metadata :foo :bpm) ; => 142

You can define your own “dynamic metadata” with define-bdef-dynamic-metadata.

(make-splits (list 0 0.25 0.5 0.75) :bdef (bdef :foo)) ; splits at the start, 25%, 50%, and 75% into the file.

(splits-from-audacity-labels "/path/to/label.txt") ; make a splits object from an Audacity labels file.

(setf (bdef-splits :my-bdef) *) ; set the :my-bdef bdef's :splits metadatum to the splits object generated from the above.

(splits-point :my-bdef 3 :start :second) ; get the start of :my-bdef's fourth split in seconds.

(pbind :instrument :playbuf
       :bufnum (bdef :my-bdef)
       :split (pwhite 0 (1- (splits-length :my-bdef))) ; pick a random split
       :embed (psplits) ; the psplits pattern yields events with :start, :end, and :dur keys to play the split specified by :split from the :splits metadatum of the bdef specified as :bufnum.
       :dur 1)

(bdef::splits-from-aubio-onsets "/path/to/audio/file.wav")

(bdef :pee "/path/to/pee.wav") ; since no BPM is listed in the filename, aubio is used to detect it (if installed)...

(bdef-metadata :pee :tempo) ; ...and it is stored in the bdef's :tempo metadatum! nice!

(bdef::splits-from-op-1-drumset "/path/to/op-1-drumset.aif") ; generates a splits by parsing the metadata in the file.

Note that any aif or aiff file will automatically be checked for OP-1 metadata, which will be parsed and stored in the splits bdef metadata key if it is found.

Currently, bdef supports SuperCollider via cl-collider as a backend. There is also basic (likely buggy) Incudine support - this will be improved later.

To write your own backend, you will need to implement the following methods on your backend’s buffer class:

• bdef-backend-supported-file-types
• bdef-backend-load
• bdef-backend-free
• bdef-length
• bdef-sample-rate
• bdef-channels
• bdef-id (optional if your backend doesn’t use buffer IDs)
• bdef-file (optional if your backend doesn’t keep track of what file a buffer was loaded from)
• bdef-frames

All other functionality is derived from those functions.

For the user’s convenience, you might also want to define methods on the bdef class for the backend’s relevant functions; see the bottom of cl-collider.lisp for an example.

• Fix the various minor/not-so-minor issues marked with “FIX” in the code.
• We have bdef-frames to get buffer data; we should have support for setting buffer data as well.
• Support for configurable pathname shortcuts. (i.e. set foo as a shortcut to /a/long/path/name/, then provide ~”foo/bar.wav”~ instead of ~”/a/long/path/name/bar.wav”~.)
• “Dynamic” splits; i.e. define a set of splits as “this region in four equal-length pieces” rather than all splits being immediately “baked” as specific points.
• Allow importing as splits from .srt (subtitle) files, .tsv (tab-separated values; this seems to be what Audacity uses, and Whisper has an option to export in this format), and .vtt (WebVTT; similar to .srt)?