An important note on GWC editing

Keyboard functions

• spacebar - starts and stops playback
• 'a' - Creates or appends "cdrdao.toc", with the currently selected (or viewed) audio segment
• 'Alt-a' - Apply sample restoration algorithm (declick),to the currently selected (or viewed) audio segment
• 'u' - increase the scale on audio display
• 'd' - decrease the scale on audio display
• 'm' - select the region between the markers
• 'r' - reset the scale on audio display to 1.0
• 's' - stop playback, and automatically highlight last 1/2 second of audio played
• 'z' - zoom in on the selected region
• B - this will toggle the marker at the start of the highlighed selection (or current view)
• E - this will toggle the marker at the end of the highlighed selection (or current view)

Documentation of the GWC drop down menus (and associated icons)

• File
  • Open - Opens an audio file for editing. Any file format that libsndfile supports may be edited.
  • Save Selection As - Saves the highlighted selection (or current view) as a new file
  • Encode Selection As MP3 - Encodes the highlighted selection (or current view) as a MP3 file
  • Encode Selection as OGG/Vorbis - Encodes the highlighted selection (or current view) as an OGG/vorbis file
  • Create cdrdao toc file As - Using the song markers, this creates a Table of Contents file for use with CDRDAO.
• Edit
  • Undo - Undoes the previous edit
  • Apply DSP Frequency Filters - Applies High Pass, Low Pass, Notch or Band Pass DSP iir filters (These are not well documented)
  • Generate Pink Noise - If you do not have a good noise sample, this will overwrite the selected region with pink and or white noise, which is a good approximation for hiss. You can then use this for your noise sample.
  • Amplify - Amplify (or attenuate) the highlighted selection (or current view)
  • Declick Strong - Remove strong clicks (pops, scratches) the highlighted selection (or current view), note this automatically attempts to detect individual clicks
  • Declick Weak - Remove weak clicks the highlighted selection (or current view), note this automatically attempts to detect individual clicks
  • Declick Manual - Apply declick algoritm to the highlighted selection (or current view). This assumes there is only one click, and it is defined by the highlighted selection or current view. Use this for cases where there are very few clicks in the audio file, or a click escapes the automatic detection but you can still hear it.
  • Decrackle - Decrackle the highlighted selection (or current view) -- use this for recordings from very dirty or damaged vinyl which may have thousands of small "clicks" per second. This will also attenuate the high frequencies somewhat.
  • Estimate - When you have a large section of audio (like a dropout, or needle bounce on a scratchy or uneven vinyl), usually greater than about 500 samples, then try this. It estimates the audio that might have been in that section. This is not a very sophisticated algorithm, and often does not improve the audio. But sometimes it does. Try it and audition the result, and "undo" if you don't like the result. You may try adjusting the start or end of the selection to see if that can improve the result.
  • Sample - Identifies the "noise only" part of the audio file. The denoise algorithm will take noise sub-samples from this highlighted selection (or current view).
  • Denoise - Applies the denoising algorithm to the highlighted selection (or current view), see Settings->denoise for options.
  • Cut - Identifies either the start or end of the audio file to be truncated. The truncation will occur when you exit GWC or open another audio file.
  • Reverb - Uses the TAP Reverb software to apply reverberation to the highlighted selection or current view. It can add a little "presence" back into the audio if the denoising seemed to take it out. For this application, I recommend your wet levels be very small indeed, try -30 Db for wet, -1 Db for Dry, and 1500 ms with the "Ambience (Thick) - HD" Reverb setting. You can of course apply more reverb if you desire it.
• View
  • Zoom Select - Expand the highlighted selection so it fills the entire window.
  • Zoom In - Zoom in by a factor of 2. The audio stays centered in the window.
  • Zoom Out - Zoom out by a factor of 2. The audio stays centered in the window.
  • View All - Display the entire audio file in the window.
  • Select All - Highlights the entire window
  • Spectral View - Displays a sonogram of the current window. This can be especially useful for locating individual clicks or pops, which show themselves as bright, vertical lines.
• Markers There are 2 kinds of markers in GWC. Editing markers, which precisely identify a location in the audio, and the highlighting function will "snap" to a marker which is less than 10 pixels from the cursor. The second type of marker is the song marker. This is used for cdrdao, and make the creation of a cdrdao toc file very simple. (Do an internet search for cdrdao for more info on that great program).
  • Toggle Beginning Marker - Sets or unsets an edit marker at the beginning of the highlighted selection (or current view).
  • Toggle Ending Marker - Sets or unsets an edit marker at the ending of the highlighted selection (or current view).
  • Clear Markers - Clears all edit markers
  • Mark Songs - Automatically place markers between songs. This uses the quiet sections between songs to place the markers
  • Move Song Marker - The song marker closest to the start of the highlighted selection is moved to the start of the highlighted selection
  • Add Song Marker - A song marker is added at the start of the highlighted selection
  • Delete Song Marker - Song markers within the highlighted selection are deleted
  • Next Song Marker - Highlights a small region around the next song marker, wraps from last to first at end of audio file.
• Settings
  • Declick
    • Weak Declick Sensitivity - Set this to something greater than 1 (1.5 is the largest you should try) to detect even weaker clicks (but you will get more "false positives"), set it lower to detect a little bit stronger clicks.
    • Strong Declick Sensitivity - This takes the value 0.75 by default. Lower values will only detect stronger clicks, higher values will detect weaker clicks.
    • FFT Weak Declick Sensitivity - Set this to something greater than 2 to detect even weaker clicks (but you will get more "false positives"), set it higher to detect a little bit stronger clicks.
    • FFT Strong Declick Sensitivity - This takes the value 5 by default. Lower values will only detect stronger clicks, higher values will detect weaker clicks.
    • Use FFT click detector - Check this box to use FFT click detection. It appears to generate fewer false positives and fewer false negatives, based on limited testing.
    • Iterate in repair clicks until all repaired - The click detection algorithm compares each candidate click against the overall sound profile in the general region of the candidate click. If the click "looks" a lot like the other sound energy (for example a cymbol crash contains a lot of the same frequencies as a click), then the candidate is ignored. But, sometimes a stronger click may mask out a weaker click because of this comparison process. So, this option causes the click detector to run on a region of the audio after any click is repaired, to re-check for more possible clicks in the absence of those clicks that were repaired. The iterate function does not yet work for FFT click detection/removal
  • Decrackle -
    • Decrackle level -
    • Decrackling window -
    • Decrackling average window -
  • Denoise
    • FFT_SIZE - You have to think about this standing on your head. (Took me a while at least). Since the samples are say, 1/44100 of a second apart, the highest frequency detectable by the FFT will always be 1/22050 of a second, regardless of the fft window size. A larger and larger fft window size allows you to detect lower and lower frequencies, and to get more of an "averaging" effect of the higher frequencies which are present in the fft window.
    • Reduction - The amount of noise you want removed from the audio, as a proportion. If you use 1.0, then all noise is removed, but unfortunately you get nasty artifacts in the resulting audio, it may get burbly or metallic sounding. In general, you should set this value as low as possible (0.3 for Blackman window, 0.5 for others), and test a few spots in the audio file (remembering to "undo" each test!!!).
    • Smoothness - For the Blackman window, it determines how the windowing function "steps along" the audio file. In general, this should be set between 5 and 11, usually just leave this at 11
    • # noise samples - The denoise algorithm sub-samples the audio data you marked for "sampling". Because noise varies from one millisecond to the next, gwc will take many sub-samples and take the "average" noise signature from those sub-samples. You should usually just leave this at 16. Don't worry if the (number of noise samples)*FFT_SIZE is greater than the sample area, gwc will just overlap the sub-samples.
    • gamma - The Ephraim-Malah and Lorber-Hoeldrich noise reduction algorithms attempt to reduce the metallic sounding artifact in denoising by holding the noise removal process very constant from one window frame to the next. Gamma is the parameter that controls this, if you set it to 0, you essentially get Weiner noise reduction on every frame, set it close to 1, and you get some weird sounding stuff. Setting it in the range of 0.5 to 0.95 usually is pretty good. Play around with this and the reduction amount f you have that metallic sound after denoising.
    • Windowing Function NEED SOME PICTURES HERE
      
      • Blackman - Try this second
      • Hybrid Blackman-Full Pass - Fastest, may work well for relatively "clean" audio files with only a little hiss.
      • Hanning-overlap-add - Usually, this is the one you want
      -
    • Noise Suppression Method
      • Weiner - A little better than Power Spectral Subtraction
      • Power Spectral Subtraction - Crudest
      • Ephram-Malah 1984 - Pretty darn good
      • Lorber & Hoeldrich - Large improvement over Ephram-Malah
  • MP3 settings
    • A bunch 'o settings here -
  • OGG settings
    • A bunch 'o settings here -
  • Miscellaneous
    • Seconds of audio preselected when "s" key is struck -
    • Seconds of audio preselected when "F7" key is struck -
    • Normalize values for declick, denoise? -
      The normalize option asks libsndfile to normalize the audio data on the interval -1.0 to 1.0, it appears to make no difference, and it really only there for testing. Set to 1 for normalization, 0 for non-normalized data.
    • Silence estimate in seconds for marking songs -
    • Log frequency in sonogram -