Snd is a sound editor modelled loosely after Emacs and an old, sorely-missed PDP-10 sound editor named Dpysnd. It can accomodate any number of sounds at once, each with any number of channels. Each channel is normally displayed in its own window, with its own cursor, edit history, and marks; each sound has a 'control panel' to try out various changes quickly, and an expression parser, used mainly during searches; there is an overall stack of 'regions' that can be browsed and edited; channels and sounds can be grouped together during editing; edits can be undone and redone without restriction; Snd can be customized and extended using C plugins, or Gnu Guile; and it's free; the code is available via anonymous ftp from ccrma-ftp.stanford.edu as pub/Lisp/snd.tar.gz.
Once compiled and loaded (see README.Snd for instructions), fire Snd up with
snd some.snd
where "some.snd" is any available sound file. You should get a window with the first .1 seconds of the sound displayed as a time domain waveform. Click the file name at the lower left to get some information about the file. Click the "f" button, and an fft window appears alongside the waveform. Drag the upper scale at the bottom of the graph and both graphs are updated as you move through the file. Drag the lower scale to zoom in or out. Drag the outer scale (the wider one) on the left to zoom the y axis in or out. The inner scale moves the y axis up and down; its main use is to move the waveform out of the way of a mix console. Click the "w" button (to unset the button) and the time domain waveform goes away. Click "play" to play the file. "sync" is more complicated -- it is used to group sounds together for simultaneous editing.
Now return to the time domain form (click "w" and "f"), and click on the graph itself. A red cursor (a big "+") appears at that point. The cursor is exactly like an Emacs cursor -- you can delete the sample at the cursor, for example, by typing control D (abbreviated in this document C-d), or move back one sample with C-b. To see all the cursor-related commands, go to the "Help" Menu, and choose "Click for Help". The cursor becomes a question mark. Move the cursor to the graph portion of the window and click, and a help window pops up, telling all about it. You can use the same sequence to get help for anything in the Snd window, even the vertical bar after the file name.
Now click and drag the mouse through some portion of the graph -- the portion dragged is highlighted in some way. When you release the mouse button, the highlighted portion is 'selected' -- it becomes a 'region' that can be deleted, pasted elsewhere (move the mouse and click the middle button), played, and operated on in many ways. You'll notice if you make some change to the data that the file name gets an asterisk, as in Emacs, and the various 'undo' and 'redo' menu options come to life. Just for laughs, cut the selection, then click the right mouse button to get the popup menu, and try Undo followed by Redo.Next type C-m -- this sets a mark at the current cursor location. C-a goes to the start of the window; C-j jumps forward to the nearest mark. Click and drag the upper horizontal mark portion to move the mark; click the triangular lower portion to play from the mark. Drag the triangular portion to play the data following the mouse.
Finally, go to the View menu and select 'Show controls'. A new portion of the Snd window is opened, containing a number of controls. Try goofing around with them while playing the sound. For the more complex cases, the button on the right side turns the option on or off. The sequence of changes you make while playing can be remembered (the 'Record' button), replayed ('Replay') or applied to the data as a kind of giant editing operation ('Apply').
File Menu
Open: open a new file (C-x C-f)
Close: close a file, flush any unsaved edits (C-x k)
Save: save current edits, overwriting previous version (C-x C-s)
Save as: save current edits under a new name, stay in current file
Revert: flush edits
Update: re-read file (data changed behind Snd's back)
View: open file read-only
Mix: mix file
New: create a new, empty file
Record: record sound producing a new file
Print: produce a Postscript version of current display
Exit: leave Snd, flushing all pending edits
Snd operates on sound files. When invoked, Snd scans its arguments for file names, and opens any it finds.
snd oboe.snd fyow.snd
If there are no arguments, Snd comes up as a bare menu bar. If a name is preceded by "-p" or "-preload", it is treated as a directory name and all sound files found in that directory are preloaded into the previous files list. To load arbitrary Snd customizations (scheme code, or a saved state file for example) precede the file name with "-l" or "-load".
snd oboe.snd -l examp.scm
opens the sound file oboe.snd and loads the Guile/Scheme code in examp.scm.
Each sound file in the argument list can include start up information about the size and location of the initial waveform display:
snd oboe.snd -s "0.0,0.25,0.75,0.3". The flag "-s" introduces the size information, followed by a string containing the values of the x position, y position, x zoom, and y zoom scrollbars.
Normally Snd adds each new sound below those currently being displayed. To position sounds horizontally (adding on the right), use the "-h" (or "-horizontal") flag, or the horizontalPanes resource. (This only works on systems with Motif 2.0 or later). Other overall layout choices include the Notebook widget (horizontalPanes: 2, -notebook), an outer scroller widget (horizontalPanes: 3, or -scroller), and separate windows for each sound (-separate). These special choices need to be set either in the resource file, or in the invocation line.
A file can be opened from the File menu via Open or View. View opens the file read-only, whereas Open will allow it to be changed. The equivalent keyboard command is C-x C-f. You can also drag a sound file's icon and drop it on top of the Snd icon or the Snd menubar. If a file cannot be written (either it was opened read-only or you don't have write permission on it), a lock appears next to the file name. Similarly, if Snd notices that the file on the disk no longer matches the original, a bomb appears (this can happen if some other program writes a sound while you are editing an earlier version of it; see also overwrite-check and corruption-time). If the variable auto-update is 1 (default is 0), Snd automatically updates any such file (this is also useful if you're editing the same file in two separate panes).
In the file selection dialog, there's a button marked 'Sound files only'. This can be set by default (see Customization). If it is set, only files that appear to be sound files are included in the directory listing. Currently Snd uses the file extension to decide whether it is a sound file -- the initial extensions are snd, wav, aiff, aif, au, aifc, voc, and wve. To add an extension to this list, use add-sound-file-extension.
Snd can handle the following file and data types:
'Linear' here means 2's complement integer. The files can have any number of channels. Data can be either big or little endian. The default output type is settable as the defaultOutputType resource. A raw data file gets its srate, chans, and data format information either from the variables raw-srate and friends (if raw-defaults == 1), or from a dialog window that pops up when such a file is opened. The files listed as "automatically translated" are decoded upon being opened, rewritten in Sun 16-bit format to a new file with an added (possibly redundant) extension .snd, and that form is the one the editor sees from then on.
Each file has its own 'pane', a horizontal section of the overall Snd screen space; within that section, each channel has a pane, and below the channels is the 'control pane', normally hidden except for the file name and 'minibuffer'. At the very bottom (or far right) of the Snd window is the lisp listener, if any (see the View menu "Open listener" option). The panes can all be independently raised and lowered by dragging the pane buttons on the right. The channel that is currently active (receiving keyboard commands) has a thin black border around the display portion. Each channel has the four scrollbars setting what portion of the data is displayed; the 'w' button (normally set) which causes the time domain waveform to be displayed; and the 'f' button (normally unset) which includes the frequency domain (FFT) display. There is a third display settable by user-provided functions; if both the 'w' and 'f' buttons are off and there is no active user-display function, you get an empty display. For each sound there is a control panel containing the file name, in parentheses if the file is actually a link, with an asterisk if there are unsaved edits, a 'minibuffer' for various kinds of text-based interactions, a 'sync' button for grouped display and edit operations, a 'unite' button (if the sound has more than one channel), and a 'play' button to play the current (edited) state of the file. Any number of files can be played at the same time. The rest of the control panel modifies how the file is played.
To open a new, empty file, use the New option.
To close a file (flushing any unsaved edits), use the File menu Close option, or C-x k. This command applies to the file that contains the active channel, or the top file in the display.
To save the current edited state of a file, use the Save option (to overwrite the old version of the file), or Save as (to write to a new file, leaving the old file unchanged). The equivalent keyboard command is C-x C-s (save). Other related keyboard commands are C-x w (save selection as file), and C-x C-w (extract and save the current channel as a file). Normally, if the new file already exists, and it is not currently being edited in Snd, it is silently overwritten. If you try to overwrite a file, and that file has active edits in a different Snd window, you'll be asked for confirmation. If you want Snd to ask before overwriting a file in any case, set the resource overwriteCheck to 1, or include the expression (set-ask-before-overwrite 1) in your Snd initialization file. If you edit a write-protected file and try to save the edits, Snd will try to save the edits in a temporary file and will post a warning that the current file could not be edited. A similar sequence can occur when opening a file on a write-protected directory where the sound file uses some compression scheme that Snd wants to translate before editing.
To undo all edits and return to the last saved state of a file, use the Revert option. The edit history is still available, so you can redo all the edits in order simply by calling Redo repeatedly.
The Print option fires up the Print dialog. You can send the currently active graph directly to a printer, or save it as an "encapsulated Postscript" file. The default name of this file is "snd.eps"; the resource that overrides this default is named 'epsFile'.
To mix files, see "Mix Files". To record a file, see "Record Files".
Finally, to exit Snd cleanly (that is, removing any temporary files, and cleaning up some system stuff), use the Exit option. Unsaved edits are silently flushed.
View Menu
Show controls: show/hide the control panel (C-x C-o, C-x C-c)
Normalize: cleanup current Snd display
Open listener: show/hide lisp listener
Channel style: separate, combined or superimposed channels
Graph style: use dots, lines, or filled polygons in the data displays
Verbose cursor: describe the current sample every time the cursor moves
Regions: a browser to examine the region stack
Files: a browser of currently and previously loaded files
Color: a browser that chooses color maps
Orientation: a browser that sets spectrogram and wavogram orientation
Groups: group editor (tracks for mixing)
Show marks: show or hide current marks
Hide axes: show or hide the x and y axes
Show y=0: show or hide the y=0 line
X axis units: x axis labelled in seconds, samples, percent of total
Error History: saved error messages
The file display can be modified in various ways. To use dots rather than connected lines, use the Dots option in the View menu's Graph style option (see also set-dot-size). Similarly, to show (or hide) the line Y = 0, use the y=0 option. The Region browser is described under Regions. To open the control panel, use Show Controls. Similarly, to open or close the lisp listener panel, use Open listener. You can also drag the associated pane button. When many files are displayed, and the pane buttons have been in use, the overall display can become a bit of a mess. To return to a state where each file and channel has a 'normal' amount of space, use the Normalize option.
The Color and Orientation options activate windows that set various aspects of the sonogram, spectrogram, and wavogram displays. There are eight or nine colormaps available along with ways to invert the maps, and scale (darken) them differently according to screen or printer characteristics. And the graphs themselves can be rotated and resized. The color variable default settings produce good output on the SGI but looks cluttered on some Linuces -- you may have to play around with the lightness scaler and so on to find reasonable values for your system.
The Popup menu's Info dialog can be left in view and updated with M-v i to reflect the currently active sound. Except for the header comment, the same information is displayed in the minibuffer when you click the file name.
The Files option fires up the file browser:
The file browser window shown above provides two lists, one of the currently active files in Snd, and the other of previously active files. The currently selected sound is highlighted. The save button saves current edits, if any; the play button plays the file; and the unlist button removes a file from the previous files list. Click a current file name, and that sound becomes the selected sound in the main Snd window. Click a previous file name, and that file is opened in Snd. The 'update' button runs through the previous files list checking for files that have been deleted or moved behind Snd's back. 'Clear' clears the previous files list. The previous files list can be preloaded via the -p switch to Snd, and the extended commands preload and preload-file-browser. By preloading your "working set" of sounds, you can save the bother of picking them up one by one from the clumsy file selection box.
The Error History option displays whatever errors have been posted during the current run of Snd. If some error gets erased too quickly from the 'minibuffer', you can recover it in this dialog.
The Regions option is described below.
Options Menu
Transform options: various transform choices
Speed style: control panel speed scrollbar interpretation
Zoom style: where to focus during zooms
Save options: save current state of options
Save state: save current state
Show stats: show disk and memory usage
The Options menu sets various preferences. The Transform Options menu applies mainly to the FFT display triggered by setting the 'f' button in the channel window. The dialog that is launched by this menu item has six sections: on the upper left is a list of available transform types; next on the right is a list of fft sizes; next is a panel of buttons that sets various display-oriented choices; the lower left panel sets the current wavelet, when relevant; next is the fft data window choice; and next to it is a graph of the current fft window; when the window has an associated parameter (sometimes known as "alpha" or "beta"), the slider beneath the window list is highlighted and can be used to choose the desired member of that family of windows.
The FFT is taken from the start (the left edge) of the current window and is updated as the window bounds change. If you'd like the fft size to reflect the current time domain window size:
(add-hook! graph-hook (lambda (snd chn y0 y1) (if (and (ffting) (= (fft-style) normal-fft)) (set-fft-size (expt 2 (ceiling (/ (log (- (right-sample) (left-sample))) (log 2.0))))))))
The fft data is scaled to fit between 0.0 and 1.0 unless the fft normalization is off. The full frequency axis is normally displayed, but the axis is "dragable" -- put the mouse on the axis and drag it either way to change the range (this is equivalent to changing the variable spectro-cutoff). You can also click on any point in the fft to get the associated fft data displayed; if verbose-cursor is on, you can drag the mouse through the fft display and the description in the minibuffer will be constantly updated.
The harmonic analysis function is normally the Fourier Transform, but others are available, including about 20 wavelet choices, and autocorrelation.
The top three buttons in the transform dialog choose between a normal fft, a sonogram, or a spectrogram. The "peaks" button affects whether peak info is displayed alongside the graph of the spectrum. The "dB" button selects between a linear and logarithmic Y (magnitude) axis. The "log freq" button makes a similar choice along the frequency axis.
The easiest way to change the colormap and graph orientation of the spectrogram, wavogram, and sonogram, is to use the Color and Orientation dialogs from the View menu. You can also use the numeric keypad, or Lisp expressions. The keypad keys are mapped to various variables as follows:
variable increase decrease
spectro-cutoff PageUp (9) PageDown (3)
spectro-hop Add (+) Subtract (-)
spectro-z-angle RightArrow (6) LeftArrow (4)
spectro-x-angle Ctrl-UpArrow (8) Ctrl-DownArrow (2)
spectro-y-angle Ctrl-RightArrow (6) Ctrl-LeftArrow (4)
spectro-z-scale UpArrow (8) DownArrow (2)
fft-size Multiply (*) Divide (/)
dot-size Delete (.) Insert (0)
You can rotate the spectrogram around the various axes by holding down the keypad and control keys. You can get arbitrarily small or large ffts with the Multiply and Divide keys. The x and y axis scalers are named spectro-x-scale and spectro-y-scale. The keypad Enter key resets all the spectrogram variables to their default values. (In Linux, use the corresponding numbered keys -- add shift to the key sequences given above). See also the Color and Orientation menu options in the View menu.
If the choice of sizes in the fft size list doesn't include the one you want, you can get arbitrarily small or large ffts with the Multiply and Divide keys. The keypad Enter key resets all the spectrogram variables to their default values. See also the fft-size variable. To interrupt a long transform-related computation, deactivate the 'f' button.
A somewhat frivolous feature, known to Snd as the
"wavogram", mimics the spectrogram in the time domain.
The same rotation commands apply to this display,
with the additional variable wavo-hop which sets the
density of the traces. To get this display M-x (set-wavo 1). It is
important in this case to get the length of each trace
correct so successive peaks more or less line up. The
trace length in samples is set by the variable wavo-trace,
or the numeric keypad + and - keys.
The Save options menu option is best explained as a part of the customization process described below. It basically sets up (or adds to) an initialization file for Snd that sets all the options to their current settings; this is intended to be similar to a .emacs file (its default name is .snd).
The Zoom style option determines the graph point that tries to remain stable in the display during an x-axis zoom. The default is to zoom onto the cursor or the beginning of the current selection if either is visible. You can also have zoom focus on the left edge, right edge, or midpoint of the current window.
Edit Menu
Undo: Undo last edit (C-x C-u or C-_)
Redo: Redo last edit (C-x C-r)
Find: Global search via find dialog (C-s, C-r)
Cut: Cut (delete) selected portion
Paste: Paste (insert) selected portion (C-y, C-x i)
Mix selection: Mix (add) selected portion (C-x q)
Play selection: Play selected portion (C-x p)
Save selection: Save selected portion as file (C-x w)
Select all: select entire file (following sync state)
Edit Envelope: Edit or view envelopes
Edit Header: Edit or view file header
Editing in Snd is modelled after Emacs in many regards. Each channel has a cursor (a big "+"), a set of marks, and a list of edits that have not yet been saved. Most operations take place at the cursor. Operations can be applied simultaneously to any other channels or sounds by using the 'sync' button. And as in Emacs, there is a notion of a region ('selection'), and a list of saved regions. Operations can be applied either to a sample, a region, a channel, a file, or any number of files at the same time. Where an operation has an obvious analog in text editing, I've tried to use the associated Emacs command. To delete the sample at the cursor, for example, use C-d.
The following sections describe how to move the cursor and the window; how to change which channel is active; how to use marks and regions; how to perform various common editing operations. It ends with a description of all the mouse and keyboard editing commands. The 'control panel' provides more complex editing operations, but has a chapter to itself.
The cursor (and its associated channel) is activated by clicking on the time domain waveform. The active channel normally has a dark edge around the graphics portion, as a sort of highlight. It is the window that receives keyboard commands. You can also move between windows with C-x o (forwards or backwards).
Any mouse click on the waveform causes the cursor to move to that point. To move the cursor from the keyboard, use:
< move cursor to sample 0 > move cursor to last sample C-< move cursor to sample 0 C-> move cursor to last sample C-a move cursor to window start C-e move cursor to window end C-b move cursor back one sample C-f move cursor ahead one sample C-n move cursor ahead one 'line' C-p move cursor back one 'line' C-v move cursor to mid-window C-i display cursor info C-j go to mark C-x j go to named mark
All keyboard commands accept numerical arguments, as in Emacs. If the argument is a float, it is multiplied by the sampling rate before being applied to the command, so C-u 2.1 C-f moves the cursor forward 2.1 seconds in the data.
The simplest way to move the window (the portion of the data in the current graph) is to drag the scrollbars with the mouse. The darker scrollbars zoom in and out; the lighter bars move the window along the x or y axis. Because sound files can be enormous, the x axis placement needs some special handling to make it useful in all cases. To move by a single windowful, click the arrows on the scrollbar. To move by smaller amounts, use the left and right arrow keys (or zoom with the up and down arrow keys); the control, shift, and meta keys are multipliers on this movement -- each key adds a factor of .5 to the multiple, so to move by .25 windows, press control, meta, left (or right) arrow. A similar mechanism can be used to zoom quickly onto a particular point; hold the keys and click the mouse in the waveform and you'll zoom an increasing amount into the data at that point.
Various keyboard commands provide much more precise control of the window bounds and placement:
C-l position window so cursor is in the middle C-x b position window so cursor is on left margin C-x f position window so cursor is on right margin [Down] zoom out, amount depends on shift, control, and meta [Up] zoom in [Left] move window left [Right] move window right C-x l position selection in mid-view C-x v position window over current selection C-x C-b set x window bounds (preceded by number of leftmost sample) C-x C-p set window size (preceded by size as numeric argument)
As in most other cases, the sample numbers (or sizes) can be floats; if the argument is not an integer, it is multiplied by the sampling rate before being applied to the command. So, C-u .1 C-x C-p makes the window display .1 seconds of data.
A mark marks a particular sample in a sound file (not a position in that file). If we mark a sample, then delete 100 samples before it, the mark follows the sample, changing its current position in the data. If we delete the sample, the mark is also deleted; a subsequent undo that returns the sample also returns its associated mark. I'm not sure this is the right thing, but it's a lot less stupid than marking a position.
Once set, a mark can be moved (redefined) by dragging the horizontal tab at the top. Click on the triangle at the bottom to play (or stop playing) from the mark; drag the triangle to play following the mouse.
A mark can be named or unnamed -- the name is displayed above the horizontal tab at the top of the window. The following keyboard commands relate to marks:
C-m place (or remove if argument negative) mark at cursor C-x / place named mark at cursor C-x C-m add named mark C-j go to mark C-x j go to named mark
The distance from the cursor to a mark can be used as a numeric argument for other commands by following C-u with C-m. Any number in-between is the number of marks to jump forward before getting the distance.
The current marks assoicated with a sound can be saved in a
file that Snd will load automatically later via
(save-marks).
A region is a (saved) portion of the sound data. Although I'm not completely consistent in this document, the word "selection" is used to refer to the currently selected (and highlighted) portion of the data; an operation such as filter-selection affects the underlying data. A "region" on the other hand, refers to the saved (copied) version of that data that can be inserted or mixed elsewhere. Regions can be defined by dragging the mouse through a portion of the data. If the mouse drags off the end of the graph, the x axis moves, in a sense dragging the data along to try to keep up with the mouse; the further away the mouse is from the display, the faster the axis moves. (One minor caveat: if you drag the mouse too quickly off the end of the graph, making a grand sweeping gesture, the last portion of the graph may be missed because the mouse updates are coalesced to some extent; move deliberately as you near the end of the sound). A region can also be defined with keyboard commands, much as in Emacs. C-[space] starts the region definition and the various cursor moving commands continue the definition.
Once defined, the copied selection is added to a stack of currently available regions (the stack size is normally 16 -- see max-regions). In many cases, the numeric argument to the command below chooses which region in that stack to use in the command. Since it can be hard to remember which region is which, especially since they are constantly on the move as the stack changes, there is a region browser, similar to the File browser, invoked from the View menu's Regions option:
The 'save' button here protects the region from deletion as the region stack grows. 'Print' produces a Postscript rendition of the current graph contents, using the default eps output name. 'play' plays the region. The graphical display shows the waveform with arrows to move around in the channels of multi-channel regions. The 'edit' button loads the region into the main editor as a temporary file. It can be edited or renamed, etc. If you save the file, the region is updated to reflect any edits you made.
The Select all menu option (in the Edit menu) uses the region mechanism as a way to save the entire current sync'd state. Besides making it easier to select an entire file (mouse event handling can make it tiresome to pick up the last few samples sometimes), this option provides a quick way to save in a temporary place the current edits applied to some file, channel, or group thereof. You can then quickly return to this state later via the region browser.
The keyboard commands that apply to regions are:
C-y paste in current selection at cursor C-[space] start keyboard-based region definition C-x a apply amplitude envelope to selection C-x c define selection from cursor to nth mark C-x i insert selection C-x n re-evaluate expression over selection C-x p play selection or region (numeric arg selects region) C-x q mix in region (float arg = scaler) C-x w save selection as file C-x x evaluate expression over selection C-x l position selection in mid-view C-x v position window over current selection
If the current selection is active (displayed somewhere in the current time domain displays), there are several functions that can edit that portion of the current sounds (that is, the edit applies to the underlying sound, not to the selection as a separate entity).
scale-selection-by args scale-selection-to args src-selection num-or-env filter-selection coeffs order call-plug-selection plug reverse-selection env-selection env save-selection name type format srate comment convolve-selection-with filename amp smooth-selection
Also if the 'selection' button is set in the transform options dialog, (or equivalently, show-selection-transform is #t or 1), the fft display, if any, displays the transform of the selected portion. Save-selection is not always the same as save-region with the region argument set to 0; a few functions, such as reverse-selection, reverse the selected data, but leave the region copy alone, so save-selection saves the reversed form, but save-region saves the original version. This can be confusing if you use C-x p to play the selection; it will actually play the original region, not the reversed form.
The current state of the undo/redo list can be viewed as a scrolled list of strings in the pane on the left of the graph (in Motif 1, there's a 'Show Edit History' menu option). If there are no current edits, it just lists the associated file name (i.e. the zero-edits state). As you edit the sound, the operations appear in the edit list window. Click on a member of the list to move to that point in the edit list (equivalent to some number of undo's or redo's). To move to a given edit point and follow the sync chain (if any), use control-click.
If Guile is loaded, the function save-edit-history saves the current edit list as a loadable program (assuming the base sounds haven't changed in the meantime). The file can be edited (it's just a text file with comments). (more to come...)
Most of these kinds of operations are accessible from the File menu. They can also be invoked from the keyboard:
C-x k close currently selected file C-x w save selection as file C-x C-d print C-x C-f open file C-x C-s save file C-x C-w save currently selected channel as file
The Print command produces a PostScript file which can be sent to directly a printer or saved for later use.
The fastest way to delete a section is to drag the mouse through it and call the Edit menu's Cut option. Any active region can be pasted or mixed in using C-x q and C-x i with a numeric argument (the region number); the current selection (region 0) can be pasted in by clicking the middle mouse button. The associated keyboard commands are:
C-d delete sample at cursor C-h delete previous sample C-k delete a 'line' -- line-size (128) samples C-w delete current selected portion C-o insert a zero sample at cursor C-x i insert region at cursor (arg = region number) C-x C-i insert file C-y paste in current selection at cursor C-x q mix in region (float arg scales mix, int arg = region number) C-x C-q mix in file
Insertion takes place just ahead of the cursor; to insert after the cursor, define a command something like:
(bind-key (char->integer #\i) 0 "(insert-region (1+ (cursor)))")
Normally each operation applies only to the currently active channel. If, however, the sound's 'sync' button is set, the operations apply to every sound or channel that also has the sync button set to the same value. If you click the 'sync' button, its value is 1 and its color is blue; C-click gives 2 and green, C-M-click gives 3 and yellow; C-M-Shift-click gives 4 and red; any other value (set via the set-syncing function) shows as a black button. The point of all this is that only those sounds that share the sync value of the current sound are considered to be sync'd to it; for example, to make a stereo selection in one file, then paste it into some other stereo file, set the sync buttons in each sound, but use different values; that way, the channels within each sound are sync'd together (giving stereo operations), but the sounds themselves are separate.
A multichannel sound also has a 'unite' button to the left of the 'sync' button. If this button is set, all channels are displayed in one graph; the x and y-axis scrollbars apply to all the channels at once, as do the 'f' and 'w' buttons; two new scrollbars appear on the right of the window; the furthest right scrollbar affects the placement of the window within the overall set of graphs, and the scrollbar on its left zooms in and out of the overall graph. For stereo files, this is user-interface overkill, but the hope is to accomodate sounds with many channels, making it easy to focus on particular portions and so on. The View menu Channel style option has the same effect but applies to all active multichannel sounds. Control-click the unite button to get superimposed channels. If the channels are not combined (the default), control-click the 'f' or 'w' button in one channel to affect all channels at once.
To get multi-channel selections, set the sync button, then define the selection (by dragging the mouse) in one channel, and the parallel portions of the other channels will also be selected.
An envelope in Snd is a list of x y break-point pairs. The x axis range is arbitrary. For example, to define a triangle curve: '(0 0 1 1 2 0). There is no (obvious) limit on the number of breakpoints. Envelopes can be defined with defvar and referred to thereafter by name. This can save much typing. Use the envelope editor to draw envelopes with the mouse.
To apply an envelope to a sound, use the extended command C-x C-a. If this command gets a numeric argument, the envelope is applied from the cursor for that many samples. Otherwise, the envelope is applied to the entire file.
C-x a apply amplitude envelope to selection C-x C-a apply amplitude envelope to channel
You can also specify an envelope name to the C-x C-a prompt.
To scale a file or selection by or to some amplitude, use the lisp functions:
scale-by args scale-to args scale-selection-by args scale-selection-to args
scale-by scales the current sync'd channels by its arguments, and scale-to scales them to its arguments (a "normalization"). The arguments in each case are either a list of floats corresponding to each successsive member of the current set of sync'd channels, or just one argument. In the latter case, scale-by uses that scaler for all its channels, and scale-to normalizes all the channels together so that the loudest reaches that amplitude (that is, (scale-to .5) when applied to a stereo file means that both channels are scaled by the same amount so that the loudest point in the file becomes .5).
Searches in Snd refer to the sound data, and are, in general, patterned after Emacs. When you type C-s or C-r, the minibuffer below the graph is activated and you are asked for the search expression. The expression is either a piece of C-like code that describes which sample satisfies the search, or the corresponding Scheme procedure. For example, to look for the next sample that is greater than .1, we could type: y > .1. The cursor then moves to the next such sample, if any. Alternatively, we could use Scheme: (lambda (y) (> y .1)). The Scheme procedure should be a function of one argument, the current sample value; it should return #t when it reaches a sample that satisfies the search criterion. Successive C-s's or C-r's repeat the search. C-x C-s can redefine the search pattern, which is also cleared by various other commands, much as in Emacs.
The search (and C-x C-x) expression C-syntax includes most of C's operators, the standard math library, as well as a few variables internal to Snd. To get at the current sample's value (the sample that the cursor is sitting on), use 'y'. 'y(n)' refers to the sample n samples from the current one. Similiarly, each of the other channels besides the current one is accessible as y1..8(n) (where the offset argument is optional); the channels are assigned to these yn functions in the order they occur in the Snd display (sounds from top to bottom, channels in order within a sound). As an example, say you want to find where two channels differ:
C-s y != y1
Or you want to find where the current channel has three consecutive zeros:
C-s (y == 0.0) && (y(1) == 0.0) && (y(2) == 0.0)
Normally, the search applies only to the current channel. To search all active files at once, use the Edit menu's find option.
The available C operators are:
+ - * / > >= < <= == !=
( ) { } += *= /= -= = || && ? : !
The available functions are:
log log10 exp cos sin abs (i.e. fabs) pow sqrt atan acos asin cosh sinh tanh fmod ceil floor y: value at cursor (y(n) also) y1..8: value ar cursor (y1(n)) in channel n (counting sounds from top)
The search commands are:
C-r find backwards C-s find forwards
See also the functions find and count-matches, and the scanning and mapping functions (Scanning Data).
The simplest changes are:
C-z set current sample to zero C-x C-z smooth data using cosine (smooth)
C-x C-x y = .1 sets the sample at the cursor to .1. To set several samples to zero, define a selection over the desired samples, then C-x x y=0. Or use C-u preceding the C-x C-x command. As with the 'find' commands, you can use Scheme syntax here: the argument to C-x C-x should be a function of one argument, the current sample value, and should return either #f (to leave the value unchanged), or the new value. The example given above could also be expressed as: C-x C-x (lambda (y) .1). To maintain local state across function calls, you can use constructs such as:
(define filt
(let ((y0 0.0))
(lambda (x)
(set! y0 (+ x (* .9 y0)))
y0)))
Then C-u 1000 C-x C-x filt. But this can only be used once. A better version might be:
(define filt
(lambda ()
(let ((y0 0.0))
(lambda (x)
(set! y0 (+ x (* .9 y0)))
y0))))
C-u 1000 C-x C-x (filt). Now a subsequent (filt) call creates a new closure (i.e. y0 starts again at 0.0).
Another way to use expressions is to make new sounds from existing
ones. Say we want to make a new sound that is the difference of
two others: make a new file (via the File menu New option, or
M-x (new)), make room in it (via C-u 1.0 C-o or whatever), then:
C-u 1.0 C-x C-x y = y1 - y2
In this particular case, a simpler way would be to scale the first by -1, then mix in the second.
Snd supports 'unlimited undo' in the sense that you can move back and forth in the list of edits without any obvious limit on how long that list can get. The data displayed is always the edited form thereof. Each editing operation extends the current edit list; each undo backs up in that list, and each redo moves forward in the list of previously un-done edits. Besides the Edit and Popup menu options, there are these keyboard commands:
C-x r redo last edit C-x u undo last edit C-x C-r redo last edit C-x C-u undo last edit C-_ undo last edit
Revert is the same as undoing all edits.
To play a sound, click the 'play' button. If the sound has more channels than your DAC(s), Snd will try to mix the extra channels into the available DAC outputs. While it is playing, you can click the button again to stop it, or click some other file's 'play' button to mix it into the current set of sounds being played. To play from a particular point, set a mark there, then click its 'play triangle' (the triangular portion below the x axis). To 'rock the reels', that is to play while dragging the mouse back and forth, drag the mark's play triangle. The Edit menu 'Play' option plays the current selection, if any. The Popup menu's 'Play' option plays the currently selected sound. And the region and file browsers provide play buttons for each of the listed regions or files. If you hold down the control key when you click 'play', the cursor follows along as the sound is played. If 'verbose cursor' is on, the time is also displayed in the minibuffer. If you stop the play in progress, the cursor remains where you stopped it, but otherwise returns to its original position. Type space (without control) to pause and continue during playback.
In Linux, the soundcard input lines are sometimes left active by default, causing an amazing amount of hum and whatnot. To try to turn these off, call the function clear-audio-inputs.
The color of the play button corresponds to the state of the playback: normal: not playing, blue: playing, green: playing and tracking with the cursor (if any), red: playback paused.
If you're getting interruptions while playing a sound (stereo 44.1 kHz sounds can be problematic), try using a very large dac buffer. The function that sets the dac buffer size is set-dac-size: M-x (set-dac-size 65536). If you're getting clicks in Linux despite a large dac-size, try setting the OSS fragment sizes by hand: set-oss-buffers(fragments,fragment-size). OSS's defaults are 16 and 12; Snd's defaults are probably 4 and 12; the larger these numbers, the more sluggish the control panel controls.
The keyboard commands for playing are:
C-q play current channel starting at the cursor C-t stop playing C-x p play region (numeric arg selects region)
In a multi-channel file, C-q plays all channels from the current channel's cursor if the sync button is on, and otherwise plays only the current channel. Except in the browsers, what is actually played depends on the control panel. Both C-q and C-x p allow overlapped plays; that is, if you type C-q several times in succession, you'll hear several simultaneous renditions of the sound. The various 'play' buttons scattered around Snd on the other hand, interrupt the current play if you click them during a run. C-g stops any current playing.
Since mixing is the most common and most useful editing operation performed on sounds, there is relatively elaborate support for it in Snd. To mix in a file, use either the File Mix menu option or the command C-x C-q. Currently the only difference between these two is that the Mix menu option tries to take the current sync state into account, whereas the C-x C-q command does not. To mix a selection, use C-x q. The mix starts at the current cursor location.
Each individual mix portion has a "mix console", a small box displayed above the waveform containing various controls. These consoles follow the sync buttons where relevant. More importantly, consoles can be edited together by collecting related mix portions into a group, then firing up the Group Editor (via the View menu Groups option). In a sense, a group can be viewed as Snd's way of implementing a more normal mixer's "tracks".
When a section or file is mixed into the current file a mix console is associated with it, each output channel getting its own console. The console is displayed at first as a row of widgets giving the input file name, the begin and end times of the mixed-in portion (click to change from seconds to samples), then three icons:
a speaker: while pushed, the input is played an 'x': click to remove the console permanently a box: click to open (or close) the console
You can drag the widgets to change the position of the mix. Once opened, each console presents a pane with an amplitude slider for each input channel, and a speed control (srate change on the input). The initial state of the console sets the speed to 1.0, and all the input amplitudes 0.0 except the channel that matches the current output channel, which is set to 1.0 (a straight mix). To return to this state at any time, click the 'amp:' label. To turn the mix off ('mute' it, set all amps to 0.0), double click the label; to return to the last settings you made, click it with control or meta down. Similarly, click the 'speed:' label to reset it to 1.0, and click it with control or meta to return to your last settings. Each time you release the mouse button (or click the amp label) counts as another 'edit' of the file, so it is usually better to use 'undo' and 'redo' in this context, rather than repeated clicks and shift-clicks. The srate scale is interpreted in the same way as the sound pane speed control -- as a float normally, but also, if you like, quantized to semitones or integer ratios.
To change the scale interpretation, set the variables mix-amp-scaler, mix-speed-scaler, and in the group editor mix-tempo-scaler; all default to 1.0 which gives a scaling range from 0 to around 12. These numbers actually scale an exponent, so (for example) if mix-amp-scaler is set to 0.5, the scale goes from 0.0 to around 3.5; similarly if mix-tempo-scaler is 0.025, its scale goes from around .95 to 1.05.
To reduce the mix console to a single letter, double click the file name. Double click the label to return to the original row of icons. Since screen space is at a premium, this minimal form of the console can reduce clutter. You can still drag the label to reposition the mix.
When any edit is performed that changes the file within the mixed portion, the affected mix consoles are removed from the display, and the only way to return to them is to undo the offending edit. That is, if you mix a portion, then cut some part of that portion, the mix will be locked in place from then on, as if you had clicked the 'x' button on the mix console.
To turn off the constant graphics updates (which can slow down old machines like mine, and which can also be annoying when you know what you're doing), set the variable movies to 0 (it is also accessible in the Group Editor).
When a sound is mixed into a file that has its sync button on, the separate channels are tied together so that as long as the sync button is on, if you move one mix console the other sync'd consoles move with it. Similarly, any speed change is reflected automatically in the other consoles; amplitude changes however, are not copied. To make the sync'd consoles independent, turn off the sync button. Once unsync'd, the consoles remain independent unless you 'undo' enough edits to return to the sync'd state.
Normally the console is lined up with its left edge at the beginning of the mixed segment. You can move the console anywhere within the segment by setting its "anchor" to the position (sample) within the mix that should correspond to the console's left edge. See mix-anchor.
To see the waveform of the sound being mixed, set the group editor "show mix waveforms" button, or set the Snd variable show-mix-waveforms to 1. The height of these graphs is set by mix-waveform-height. The mix waveform color is set by the resource mixWaveColor.
To tie together an arbitrary collection of mix consoles, use the 'group' buttons in the upper right corner of the mixer. Any mixers that share a button can be changed in parallel using the Group Editor (the Groups option under the View menu). The group editor gives amplitude control over each output channel that has a mix that is a member of the group. The speed control affects the sampling rate, and the tempo control affects the spacing of the individual members of the group. The play button plays just the group members.
The group editor envelope fields use the same syntax as other Snd envelopes: '(0 0 1 1) is a ramp, for example, and '(0 0 1 1 2 0) is a sort of pyramid. The envelope data in a given field only takes effect after the field has been "activated" (normally by a carriage return). The tempo envelope is interpreted in terms of the current group times using the placement of the given mix within the group to find the parallel place on the (arbitrary) tempo envelope x axis. The value of the envelope at that point becomes another tempo multiplier; that is, these envelopes are not time maps (perhaps someday!). Since a tempo value of 2 makes things happen twice as slowly (it essentially multiplies begin times), the tempo envelope causes things to go faster as it gets closer to zero -- weird! The speed envelope changes the overall speed of a given segment, but doesn't change within that segment.
The underlying mix consoles have independent envelopes, but my original plan to include envelope text fields in the consoles made them too big, cluttered, and unwieldy; an attempt to make each console a scrollable (or paned) window failed due to unfortunate limitations in Motif. These could be overcome, but would require much more programming than I think the issue is worth. I'm interested in better ideas, if anyone has any. One idea I'm goofing around with makes the mix matrix entries visicalc-like expressions.
To move the cursor from one mix to the next, in the same manner as C-j moves through marks, use C-x C-j.
Since a mix is viewed in Snd as just another edit which you can undo and redo, and since one mix console can be participating in any number of groups, and since any number of files can have these grouped mixes active at once, and since you can ungroup a mix, then undo the related edit, it becomes a bit of a pain to describe what each sequence of actions will do. In general, think of a group as simply another kind of 'sync' button. The Group Editor itself does not follow the undo/redo chains (edits in Snd are considered to be local to each channel, but the group editor affects all mixes together, possibly across many channels). Each change to a group control immediately affects all mixes associated with that group, in each case as a single edit of each participating channel. Each change to the mix console takes into account the state of the groups that it is a member of. Undo of any edit simply undoes the edit -- the mix console will back up one state, but if the group controls have changed in the meantime, the two are now out of sync. Similarly, if you remove a mix from a group, the group's effect upon the mix is removed as well. If it joins a group with current state, that causes an immediate change in the mixed portion's state. The tempo control, however, only affects mixes that are under its control when it is changed (does that make sense? -- the idea is that you might want to leave a group briefly to fix a local mix position or speed, then rejoin the group, but in that process you don't want the tempo control to cause the mix to jump about randomly in time). Similar complications affect the group envelopes. You can choose whether the envelope is applied over the entire output duration, or just the group duration; in the latter case, you need to be aware that adding or removing mixes can change that duration, changing the effect of the envelope. Group state changes by themselves are not edits (i.e. if a group has no members, undo and redo will not notice that you've been fiddling with its amp scalers). If a mix is participating in several groups, all associated groups affect the mix together (group states are multiplied, and envelopes are "melded" into something resembling a gathering of the various envelopes). If a mix is synced to others, and joins or leaves a group, all the synced mixes follow its lead. All of this can be summarized: groups act like the sync buttons. And if it's confusing, consider the alternatives; or better, decide once and for all which mixes are grouped together, and use only one group per mix. Then everything should be "intuitive".
As in Emacs, C-x ( begins keyboard macro definition, C-x ) ends it, and C-x e executes the last keyboard macro. Unlike Emacs, C-x C-e prompts for a name for the last defined macro; M-x name invokes the macro "name", the function save-macros saves all current named macros, and save-macro saves one (named) macro in the initialization file (normally .snd). See the lisp section for more details.
To change the sound file's header or data format, use the File or Edit menu Save as option. Choose the header type you want, then the data format, (the data format list will change depending on the header choice). The File version saves the current selected file. The Edit version saves the currently selected region.
Normally Snd keeps the current cursor, and therefore most actions you can perform on the current channel, within the current channel's bounds. The easiest way to extend the file is to pad it with zeros. Go to the end of the file, via the C-> command, then use the C-u command with a float argument, giving the number of seconds to add to the end of the file, followed by C-o (insert zeros): for example, to add a second, C-> C-u 1.0 C-o. The same sequence can be used to add silence to the start of a file.
To make a recording, choose "Record" from the File menu. A window opens with the various recording controls (SGI and Linux shown below):
The top three panes display the status of the input and output lines. If a channel is active, its meter will glow yellow. If some signal clips during recording, the meter will flash red. The numbers below the channel buttons indicate the signal maximum since it was last reset. The sliders underneath the meters scale the audio data in various ways before it is mixed into the output. The vertical sliders on the right scale the in-coming signals before the meter, and the output signal before it gets to the speaker (these are needed to avoid clipping on input, and to set the 'monitor' volume of the output independent of the output file volume).
The fourth pane has information about the current output file (its name and so on), and the layout of the window. The buttons on the right can be used to open and close panes painlessly. If the button is not square (a diamond on the SGI), the underlying audio hardware can't handle input from that device at the same time as it reads other "radio" button devices. So, in that case, opening the panel via the button also turns off the other incompatible device. The 'autoload' button, if set, causes the recorded file to be loaded automatically into Snd.
The fifth pane contains a history of whatever the recorder thought worth reporting. The duration field gives the current output file's duration. The bottom row of buttons dismiss the window, start recording, cancel the current take, and provide some help. There's also a slider on the far right that controls the speaker output volume (independent of the output file volume).
To make a recording, choose the inputs and outputs you want; for example, to record channel A from the microphone to channel A of the output file, click the Microphone panel's A button and the Output panel's A button. Then when you're ready to go, click the Record button. Click it again to finish the recording. The default is to provide up to two channels of output; if you want to write output files with more channels, set the recorder-out-chans variable before invoking the recorder.
If the record window's VU meters are too big (or too small) for your screen, you can fool around with the variable vu-size which defaults to 1.0. Similarly the variable vu-font-size tries to change the size of the numbers on the label, and vu-font chooses the family name of the font used (normally "courier").
The description above is aimed at the SGI Indy; in other systems, the layout may be different. In general, each separate pane reflects an independent digital data stream; the vertical sliders on the right reflect various hardware gains or tone controls that are applied before the signal gets to the ADC, or in the case of output, after it has been read in, scaled, and mixed into the output; in Linux the actual meaning of some of these sliders is an industry secret, so some experimentation may be needed. The basic signal flow can be viewed as coming in on the right of the input panels, scaled in the sound card hardware by the vertical sliders, passed through the ADC (if any), reflected in the VU meters, then scaled by the horizontal sliders of the input panel, mixed into the output signal, scaled by its horizontal sliders, reflected in its VU meters, then written to the output file, while another copy passes out the right side of the output panel to the speakers scaled by the output (vertical) sliders. The bizarre icon next to the microphone in Linux is supposed to mean "line-in".
Sometimes it is more convenient to have the recorder start writing the output file data only after some threshold amplitude has been reached. The scale bar in the lower right marked "trigger:" sets this value. If it is 0.0 (the default), the recording begins as soon as you push the "Record" button. Otherwise, the recording begins after you push "Record" and after the triggered value is noticed in the output (modulo buffer crossings -- sometimes you'll get up to a buffer's worth of whatever preceded that value). In this case, the button is labelled "Triggered Record".
If there are so many input and output channels active that more than eight sliders would be needed in a given pane, the recorder tries to conserve tight screen space by showing only the direct cases, chan0 to chan0, chan1 to chan1, and so on. It also posts a box of buttons to the left of the meters, each button representing a slider. If the button is on (green), the corresponding slider is displayed. Press one of the buttons to toggle whether its slider is displayed or hidden. It is also possible to drag the mouse through the box of buttons, setting (button 1) or unsetting (button 2) buttons as you go. Click in the corner (the '/' label) to set all (button 1) or unset all (button 2). Similarly, click above a column or to the left of a row to set (button 1) or unset (button 2) the entire column or row. Control-click in the corner to return to the default settings.
The Edit Envelope dialog (under the Edit menu) fires up a window for viewing and editing envelopes. The dialog has a display showing either the envelope currently being edited or a panorama of all currently loaded envelopes. The current envelope can be edited with the mouse: click at some spot in the graph to place a new breakpoint, drag an existing breakpoint to change its position, and click an existing breakpoint to delete it. The Undo and Redo buttons can be used to move around in the list of envelope edits; the current state of the envelope can be saved with the 'save' button, or printed with 'print'.
Envelopes can be defined using defvar, and loaded from a separate file of envelope definitions via load. For example, the file:
(defvar ramp '(0 0 1 1)) (defvar pyramid '(0 0 1 1 2 0))
defines two envelopes that can be used in Snd wherever an envelope is needed (e.g. C-x C-a). You can also define a new envelope in the dialog's text field; '(0 0 1 1) followed by return fires up a ramp as a new envelope.
In the overall view of envelopes, click an envelope, or click its name in the scrolled list on the left to select it; click the selected envelope to load it into the editor portion, clearing out whatever was previously there. To load an exisiting envelope into the editor, you can also type its name in the text field; to give a name to the envelope as it is currently defined in the graph viewer, type its name in this field, then either push return or the 'save' button.
Once you have an envelope in the editor, it can be applied to the currently active sounds via the 'Apply' or 'Undo&Apply' buttons; the latter first tries to undo the previous edit, then applies the envelope. The envelope can be applied to the amplitude, the spectrum, or the sampling rate. The choice is made via the three buttons marked 'amp', 'flt', and 'src'. The filter order is the variable filter-env-order which defaults to 40. To interrupt the application of the envelope, click the 'Stop' button. To apply the changes to the current selection, rather than the current sound, set the 'selection' button.
The two toggle buttons at the lower right choose whether to show a light-colored version of the currently active sound (the 'wave' button), and whether to clip mouse movement at the current y axis bounds (the 'clip' button). The 'linear' and 'exp' buttons choose the type of connecting lines, and the 'exp base' slider at the bottom sets the 'base' of the exponential curves, just as in CLM. If the envelope is being treated as a spectrum ('flt' is selected), the 'wave' button shows the actual frequency response of the filter that will be applied to the waveform by the 'apply' buttons. Increase the filter-env-order to improve the fit. In this case, the X axis goes from 0 Hz to half the sampling rate, labelled as "1.0".
The Edit menu's Edit Header option fires up a dialog to edit, add, or remove the sound's header. No change is made to the actual sound data; the new header is blindly written out; any unsaved edits are ignored. If you specify 'raw' as the type, any existing header is removed. This dialog is aimed at adding or removing an entire header, or editing the header comments; anything else is obviously dangerous. If you don't mess with the data location, it will be updated to reflect any header changes; that is, unless you intervene, the resultant header will be syntactically correct. After writing the new header, you should either close or update the associated sound.
At any time you can save the current
state of Snd by calling (save-state name) where name
is a file name. You can start in this state by calling Snd
with this file name and the "-l" switch: snd -l name.
This file is in exactly the same format as the initialization
file, and can be edited, renamed, or
whatever. To load such a file after startup, (load name) (C-x C-l).
Currently, mix consoles are not saved,
but the associated edits are.
Due to the quantized nature of the y-axis position scroller, a tiny signal that is not centered on 0 can be a pain to position in the window. Use the set-y-bounds function to set the y axis bounds to some small number, then use the position scroller to find the signal. For example, if your signal is a very soft recording setting only the lowest two bits of a 16 bit signal but with DC offset due to the recording conditions, M-x (set-y-axis -.01 .01) and try the scroller.
C-g at any point aborts the current keyboard command sequence, as does any mouse click. C-g can also interrupt some long computations (search, eval expression, and the various envelope applications).
C-u introduces a numeric argument. Besides the integer and float cases mentioned several times above, you can also use marks to set the argument. If the (optional) number after C-u is followed by C-m, the resultant number passed to the command is the distance (in samples) from the cursor to the n-th successive mark. That is C-u C-m C-f is the same as C-j.
C-x introduces an 'extended command'. It can be preceded by a numeric argument or aborted with C-g. C-x halts any on-going region definition.
C-x d prompts for the temporary directory name. Snd saves some editing operations in temporary files, rather than in-core buffers. If the temporary directory has been set (in various ways), it is used; otherwise Snd looks for the TMPDIR environment variable; if it is not found, or is null, the directory /var/tmp is used.
C-x C-c closes the control panel. C-x C-o opens the panel.
C-x C-x evaluates an expression; C-x C-n re-evaluates it.
To change a key binding, use bind-key.
As in Emacs or Tcsh, the Tab key in a text field invokes a context-sensitive completion function that tries to figure out what the rest of the text probably should be. If it finds no matches, the text flashes red; if it finds multiple matches and can't extend the current text, it flashes green, and (eventually) a list of possible completions appears in the help window. If there is no completion routine active, Tab is a no-op (i.e. it doesn't activate the "next" widget, as in the normal Motif program).
There are two types of keyboard commands in Snd: those associated with the Meta ('Alt') key, used as mouse equivalents for various menu options, and all the rest. The menu-related commands are shown underlined in the menubar. You type the underlined letter with Meta, then just the letter (no Meta) of the option you want. The menu-related commands are:
File Menu Edit Menu View Menu Options Menu Help Menu
M-f o: Open M-e u: Undo M-v s: Show controls M-o t: Transform Options Click for help
M-f c: Close M-e r: Redo M-v n: Normalize Speed style Overview
M-f s: Save M-e f: Find Open listener Focus style FFT
M-f a: Save as M-e c: Cut M-v c: Channel style M-o a: Save options Find
M-f r: Revert M-e p: Paste M-v d: Graph style Save State Undo and Redo
M-f m: Mix M-e m: Mix Selection Verbose cursor Show stats Sync
M-f u: Update M-e p: Play Selection M-v r: Regions Speed
M-f n: New M-e s: Save Selection M-v b: Files Expand
Record Select all Color Reverb
M-f v: View Edit Envelope M-v o: Orientation Contrast
M-f p: Print Edit Header Groups Envelope
M-f x: Exit M-v m: Show marks Marks
Hide axes Mixing
M-v y: Show y=0 Formats
Hide consoles Customization
X axis units Recording
Error History News
The rest of the keyboard commands refer to the currently active graph, and any graphs that are currently sync'd to it. In general, commands are analogous to Emacs where such an analogy exists; in other cases, an extended command followed by a key with control affects the current channel whereas the same key without control affects the selection. Case is not significant.
Most of the commands accept a numeric argument which can be either an integer or a float; an integer causes the command to be repeated that many times; the float is usually multiplied by the sound's sampling rate, then applied that many times. So, for example, C-u 1.0 C-f causes the cursor to move ahead one second in the sound. The argument can be negative to reverse the sense of the command, as in Emacs. It is often useful to use the distance to a mark as a count. If the C-u n command is followed by C-m, the 'n' determines how many marks to move to, then the count returned is the distance from the cursor to that mark. For example, C-u C-m returns the number of samples between the cursor and the next mark, so C-u C-m C-z zeros all the samples between the cursor and the next mark.
The commands are:
[Down] zoom out
[Up] zoom in
[Left] move window left
[Right] move window right
[Home] update current file (re-read data from disk)
[Space] cancel and deselect selection, return to pre-select position
if playing, pause or continue playback
<: move cursor to sample 0
>: move cursor to last sample
C-<: move cursor to sample 0
C->: move cursor to last sample
C-a: move cursor to window start
C-b: move cursor back one sample
C-d: delete sample at cursor
C-e: move cursor to window end
C-f: move cursor ahead one sample
C-g: abort current command
C-h: delete previous sample
C-i: display cursor info
C-j: go to mark
C-k: delete one line's worth of samples
C-l: position window so cursor is in the middle
C-m: place (or remove) mark at cursor location
C-n: move cursor ahead one 'line'
C-o: insert one zero sample at cursor
C-p: move cursor back one 'line'
C-q: play current channel(s) starting at cursor
C-r: repeat last search backwards
C-s: search according to an expression
C-t: stop playing
C-u: start count (numerical argument) specification
C-v: move cursor to mid-window
C-w: delete current region
C-x: start extended command (see below)
C-y: paste in region
C-z: set current sample to 0.0
C-_: undo last edit
C-[Space]: start region definition (arg<0 => cancel and deselect)
The extended commands (preceded by C-x) are:
a: apply envelope to selection
b: position window so cursor is on left margin
c: define selection from cursor to nth mark
d: get temp dir name
e: execute keyboard macro
f: position window so cursor is on right margin
i: insert region
j: go to named mark
k: close file
l: position selection in mid-view
n: re-evaluate expression over selection
o: move to next or previous graph
p: play selection or region n
q: mix in region
r: redo last undone edit
u: undo last edit
v: position window over current selection
w: save selection as file
x: evaluate expression over selection
z: smooth selection
/: place named mark
(: begin keyboard macro definition
): end keyboard macro definition
C-a: apply amplitude envelope
C-b: set x window bounds (preceded by first sample number)
C-c: hide ("close") controls
C-d: print
C-f: open file
C-g: abort command
C-i: insert file
C-j: go to mix console
C-l: load scm file
C-m: add named mark
C-n: re-evaluate expression
C-o: show ("open") controls
C-p: set (time domain display) window size (preceded by size)
C-q: mix in file
C-r: redo last undone edit
C-s: save file
C-u: undo last edit
C-v: set (time domain display) window size as percentage of total
C-w: save current channel in file
C-x: evaluate expression
C-z: smooth using cosine
The control panel is the portion of each sound's display beneath the channel graphs. It is normally hidden, except for the upper portion that shows the file name, the sync and play buttons, and the minibuffer (the portion in between that opens into a text window when activated).
The controls are: amp, speed, expand, contrast, reverb, and filter.
'Speed' here refers to the rate at which the sound data is consumed during playback. Another term might be 'srate'. Snd normally uses linear interpolation to perform the speed change; to use sinc interpolation instead, use src or the envelope editor. The arrow button on the right determines the direction we move through the data. The scroll bar position is normally interpreted as a float between .05 and 20. The Options Speed Style menu (or the speed-style variable) can change this to use semitones (actually microtones) or just-intonation ratios. The number of equal divisions to the octave in the semitone case is set by the variable speed-tones (normally 12).
'Expand' refers to a kind of granular synthesis used to change the tempo of events in the sound without changing pitch. Successive short slices of the file are overlapped with the difference in size between the input and output hops (between successive slices) giving the change in tempo. This doesn't work in all files -- it sometimes sounds like execrable reverb or is too buzzy -- but it certainly is more robust than the phase vocoder approach to the same problem. The expander is on only if the expand button is set.
The reverberator is a version of Michael McNabb's Nrev. In addition to the controls in the control pane, you can set the reverb feedback gain and the coefficient of the lowpass filter in the allpass bank (see below). The reverb is on only if the reverb button is set. The reverb length field takes effect only when the reverb is set up (when the DAC is started by clicking 'play' when nothing else is being played).
'Contrast enhancement' is my name for a somewhat weird waveshaper or compander. It phase-modulates a sound, which can in some cases make it sound sharper or brighter. For softer sounds, it causes only an amplitude change. To scale a soft sound up before being 'contrasted', use the variable contrast-amp. The function maxamp returns the channel's maximum amplitude, so the inverse of that is a good first guess for contrast-amp. Contrast is on only if the contrast button is set.
The filter is an arbitrary (even) order FIR filter specified by giving the frequency response envelope (the envelope is assumed to go from 0 to half the sampling rate) and filter order in the text windows provided. If you raise the control pane from its default height, a graph is revealed beneath the filter text field; this is an envelope editor like the envelope editor dialog but specialized for filtering. The actual frequency response is displayed in blue. The filter is on only if the filter button is set.
There are many variables that reflect or control the panel's various sliders; each also has a default or initial value. The reverb and expand functions also have several aspects that aren't brought out to sliders or buttons on the panel, but that can be accessed through these variables. See the lisp section for details.
The 's' (for 'Save') button saves the current control panel state for a subsequent 'Restore'. The 'r' (for 'Restore') button returns the control panel to the state at the time of the last save, or the initial state if there has been no save.
Except for the the reverb length setting, the controls can be changed as the sound plays. You can record the (non-filter-related) changes you make by setting the 'Record' button, and replay the new version later with 'Replay'. The latter actually takes effect only when play is set in motion. 'Replay' and 'Record' can be set at the same time, allowing you to add more changes or override previous ones. To take all the current recorded changes and turn them into a giant 'edit' of the file, click the 'Apply' button. Apply may change the length of the file; for example, if reverb is on, the reverb decay length is added onto the end. Once Apply has taken effect, the controls section is reset to its clean state (so a subsequent 'play' plays the unmodified newly edited version). It is possible to use 'Apply' over the current selection, or over the current channel (rather than the entire sound); use C-x p (for the selection) or C-q (for the entire channel), then 'Apply'. In the selection case, this only works if the selection does not span several sounds; the new data is truncated at the bounds of the current selection. (This may be changed soon).
Snd's control panel is not intended to replace mixers or CLM -- I view it as a quick-and-dirty 'cartooning' facility. After finding useful settings, you'll want to save the recorded changes and run them through CLM or some other hi-fi signal processing system.
The keyboard commands associated with the control panel are:
C-x C-o show ("open") control panel
C-x C-c hide ("close") control panel
See also call-apply.
See extsnd.html.