Usage Notes

Table of Contents


Making a Backup of an Audio CD

A dialog box can be called from the menu item [Tools] > [Read CD-DA tracks] on the main menu. On that dialog box, you can choose and import audio tracks from an audio CD to your computer. The dialog box also allows you to make a backup of an audio CD on your computer. A backup is the complete set of audio tracks copied from an audio CD, allowing you to access the tracks anytime without using the physical disc.

To make a backup, set the disc in your drive, click the "Whole Disc Backup" button and enter a name for the backup. In response, the app makes two files. One is a text file to store the disc's layout information in the Cue Sheet format and the other is an audio file to store the disc's sound. The sound will be encoded with the FLAC codec.

After a successful making of the backup, you can get the disc out of the drive. To open the backup on this dialog box, you have to change the current drive to the one named "Virtual CD-DA Drive". Then click the disc ejection button on the dialog box and choose the backup you just made. The tracks should be listed up on the dialog box as if you had set the disc in the drive again.

Because a backup is a pair of files each in a format known to developers, you can use other audio players to play your backups, but some players might not provide a way to choose a track to play if one single audio file has all sound for all tracks.

Technically, the app backs up only one session which is the one the first audio track belongs to and ignores any other sessions. In almost all cases, all audio tracks on the disc will be backed up because a typical audio CD does not have multiple audio sessions. Also, the app skips over the starting gap of the session by two seconds because that part may not be allowed to access. Thus the audio frame located at 00:02:00 MSF and the ones after that will be copied to the backup.

A cue sheet is a text file and can be viewed or edited with a text editor. If you should change the name of a backup, then in addition to renaming the backup's file pair, you will need to edit the cue sheet to update the name of the audio file appearing in it.

They are protected by the law

The reason why you can copy commercial audio CDs without difficulty is that they are not guarded by a copy protection mechanism. But this does not mean that they are neither copyrighted nor protected by the law. You should not open your backups to the public on the internet without making arrangements for legality.

Making a physical copy from a backup

If you want to make a physical copy of an audio CD with this app, you need to first make a backup of that CD as explained above. Then invoke a dialog box with the menu item [Tools] > [Write CD-DA Tracks]. On that dialog box, open the cue sheet of the backup, set a blank CD-R or CD-RW disc in your drive and then click the "Next" button. This causes another dialog box to appear. On that dialog box, click the "Start" button to start disc writing.

You might want to use a cue sheet that was generated by some other application or one that was composed from scratch. Note that in that case, the audio files the tracks refer to in the cue sheet must be either in the FLAC format or in an uncompressed PCM format in order for the app to load that cue sheet successfully.

Storing CD-TEXT information

In a cue sheet, the TITLE, PERFORMER and SONGWRITER commands can be used to provide information about the disc and the tracks. Such supplementary information is stored on a disc in the form of CD-TEXT data packs.

The CD-TEXT specification allows three character sets: 7-bit ASCII, Latin-1 and Shift-JIS. The app encodes your text with a suitable character set, choosing from the three. The specification is so legacy that you have no way to leave the text in Unicode.


Making a Set of MP3 Files from an Audio CD Backup

To make a set of MP3 files from an audio CD backup, first you need to open the backup and load the tracks into the app. Then run a batch processing script to make multiple MP3 files at once from the tracks open in the app. There are some scripts that come with the app and you can use one of them to make MP3 files.

To open a backup and load the tracks, you can use the dialog box which can be invoked with the menu item [Tools] > [Read CD-DA Tracks] on the main menu. Be sure to set the current drive to "Virtual CD-DA drive" to access backups rather than real discs. You can choose all of the tracks or some of them. Having some tracks selected, click the "Read In" button. The dialog box will close and waveform graphs will be created for the chosen tracks.

Then invoke another dialog box with the menu item [Tools] > [Run Script]. You can use the script ExportMP3.js to make MP3 files. Click the "Select" button on the dialog box and choose the name of that script from the appearing menu. This will cause the name of that script to be set in the edit box, with the /? option appended.

Just delete the /? option and run the script with the "Run" button. This will cause the script to make a folder named "script_output" in your Music folder and then MP3 files in it. After a successful run of the script, you can change the name of the folder or delete the folder after transferring the files to some device such as your phone.

The script can take some options. To specify the sound quality, use the /kbps option.

  • ExportMP3.js /kbps:192

Running the script with the /kbps option and its value as shown above will cause the script to generate MP3 files containing audio streams the bitrate of which is 192 kbps for stereo sounds. The default bitrate is 128 kbps.

To get information about the options, run the script with the /? option.

  • ExportMP3.js /?

Running the script with the /? option will cause its usage to be displayed on the dialog box.

Internal Commands

To change the size of the font the dialog box uses to show messages from the running script, you can use the internal command "font" with a size argument appended. For example, typing in "font -15" on the edit box followed by a click on the "Run" button will cause the font size to be set to 15 plus the internal leading of the font. Specifying a positive value causes the font size to be set to that value without adjustment.

To view the code of a script, you can use the internal command "dump". To view the code of ExportMP3.js, type in "dump ExportMP3.js" on the edit box and click the "Run" button. You might want to move the displayed text to a text editor via the clipboard to get better visibility.


How to Select a Part of a Sound

A new waveform graph is created inside the main window every time you open a sound file. Each waveform graph presents the Start and Finish markers to let you select a part of the sound that it shows.

The Start and Finish markers are each displayed as a vertical line. The selected region, which is the region between the two markers, is displayed with highlight colors.

The waveform graph can present a shortcut menu that provides a way to move either of the markers to a different time position. You invoke that menu with a right click (or secondary-button click) on the graph. When doing that, you can adjust the position by dragging the mouse before releasing the button. Once you select the appropriate item from the menu, the corresponding marker will change its time position to the location of the action.

If the part you want to select is wholly visible on the graph, you can select it simply by mouse dragging.

What purposes the markers are used for

You use the Start and Finish markers both to tell the app, for example,

  • which part should be cut off and/or copied to the clipboard
  • which part an effect should be applied to
  • which part should be exported as MP3 audio

You use the Start marker alone to tell the app, for example,

  • which time position the sound clip on the clipboard should be pasted at
  • which time position the spectrum view should set its initial view to

Creating MP3 Files

In the case of this app, the way you create MP3 files is different from the way you save uncompressed WAV files.

To create an MP3 file, you first invoke the encoder dialog box which provides the way to set the bitrate and other encoding options. The dialog box can be invoked by selecting the appropriate item from the encoder list [Edit] > [Compression] located on the main menu. There is also a toolbar button that shows the same encoder list. The toolbar is located on the top side of the main window.

By default, it is the sound in the selected region that the app converts to MP3. The selected region is the region between the Start marker and Finish marker. You can also have the app convert the entire sound by specifying that alternative behavior on the encoder dialog box.

If you are using Windows XP and the encoder dialog box does not present the full bitrate list, then you'll have to install an updated version of Windows Media Player.


Creating Opus Files

The Opus encoder can be invoked by selecting the menu item [Edit] > [Compression] > [Opus], or by using a toolbar button that has the same functionality. If you want to export the entire sound, check the item "Encode entire sound" on the encoder dialog box. Leaving this option unchecked makes only the selected region to be exported.

One difference between the Opus codec and other major codecs is that Opus internally manages audio data only at 48 kHz. It leaves 44.1 kHz out of sight.

If you pass the encoder a sound at any sampling rate other than 48 kHz, the encoder will resample it at 48 kHz. Most playback devices have the capability of playing sounds at 48 kHz without any severe quality loss.

The sampling rate you can set on the encoder dialog box does not change the behavior of the encoder in any way, but may change the behavior of an application when it decodes the sound. Applications are allowed to ignore this setting but some of them might decode the sound at the sampling rate you have specified. Most players would ignore this setting and simply play the sound at 48 kHz.

The encoder offers two types of VBR encoding. One is the regular type shown simply as "VBR" on the encoder dialog box, and the other is the type shown as "constrained VBR" which resembles CBR in a sense. If you prefer quality consistency throughout the encoded sound, the regular VBR encoding is for you. If you prefer size predictability for a VBR stream that is close in size to a CBR alternative, the "constrained VBR" encoding is for you.

With the "constrained VBR" encoding, the encoder controls the range of the bitrate fluctuation so that the resulting stream will be a stream, the average bitrate of which is roughly equal to the target bitrate. This means that it will be even in size with a CBR alternative but would be better in quality.


Notes on the Click Removal Effect

This audio effect removes transient noises or click noises from a sound. Typically such a type of noises are found in sounds from analog discs.

The effector finds out the time span of each transient noise and manipulates them each separately. They are usually less than 1 millisecond in length but can be longer. The effector replaces each with a piece of machine-calculated artificial waveform and leaves the other part of the sound untouched. Processing in this way keeps the overall quality of the sound unchanged.

The effector processes the sound on a frame-by-frame basis. Each frame is processed using a certain type of filter that adapts to only that specific frame. The filter is used for both noise detection and repair purposes. Finding a desired filter for a frame can be a time-consuming task and it takes a longer time to process with a longer filter. Because of these, it is a good idea to begin with the shortest filter. If the time span of a noise is short enough compared with the filter and is isolated enough from its neighbors, the attempt of removing that noise would result in success. If the noise is too long but is isolated enough, using a longer filter may yield a better result.

You can choose the length of the frame but the effect of that is not really obvious. The first choice should be the middle length setting. Changing to the short frame setting might be able to reduce false-positive detection errors. Changing to the long frame setting might be able to reduce false-negative detection errors.

  • False-positive detection errors can alter transient sounds, such as drum beats, sometimes making their intensities weaker.
  • False-negative detection errors can leave audible noises untouched.

Retrying removal

If you find the effector has passed over a sounding noise, you can let the effector know where it is. Just make a selection with the Start and Finish markers so that the noise will be located at the middle of them. The length of the selection needs to be as short as one frame or so. Then apply the effect again. The effector will omit the first detection pass which might have caused the previous passing-over and it might notice that noise this time.

Example of click noise removal
Before removal
After removal

Notes on the Hiss Reduction Effect

This audio effect reduces background noise in a sound. The target background noise in your sound is assumed to be mostly stationary. In the typical case, the noise is one of the following two or a mix of these:

  • Cassette tape hiss
  • Broad band noise added to broadcast radio wave while transmission through the air

Note that this audio effect is not good for reduction of wind blows, crowd noise, or any other types of noise that cannot be taken as stationary.

You should always check the resulting sound using headphones because applying this kind of effect usually damages the sound somewhat. You might want to leave the sound untouched if you feel the loss of quality is too heavy.

Preparing a noise sample

For the effector to work, you need to prepare a noise sample and let the effector know where it is. The first thing you have to do is find a silent region in the sound that contains only the target background noise, and then select that part with the Start and Finish markers. You can place it on the clipboard if you want to apply the effect to some limited part of the sound.

The length of a noise sample can be any, but should be longer than 0.5 second and no longer than a few seconds. A noise sample of one second would work well.

The effector analyzes the given noise sample and makes an estimation to shape a noise floor based on the analysis. Then the effector performs noise reduction on the given sound using that estimated noise floor.

To precisely determine the starting and ending boundaries of your noise sample, you might want to magnify the graph vertically. This can be done using the key combination of Ctrl+ Up (and Ctrl+ Down for reverse magnification), or you might prefer the toolbar button that functions the same. The toolbar is located on the top side of the main window.

To check the spectral characteristics of your noise sample, you can use the spectrum view. Right-click within your noise sample on the graph, then select "Spectrum" from the shortcut menu that pops up at the clicking location.

The effect settings

Some of the effect settings might need to be explained.

The effector processes the given sound on a frame-by-frame basis. Each frame contains a short segment of sound. FFT processing is performed for each frame separately.

Name Description
Frame Length The Medium setting is good for both speed and music. The Short setting is for speech. The Long setting is for music but not for pieces in which drum beats, guitar picks, piano strokes, or any other striking onsets frequently appear. It might be good for orchestral pieces, classical pieces, or other similar ones.
Frame Move Two successive frames have to be overlapped to make a smooth waveform without discontinuity. The setting of 1/4 frame advancing would work well in most cases. Choosing either of the other two would not make a big difference.
Attenuation The effector tries to attenuate the noise by this amount. Too much attenuation can result in too much quality loss. In general, a choice within the range between -1 and -15 dB would yield a good result.
Exclude outstanding frequencies The estimated noise floor is further shaped so that it does not contain spike-looking outstanding components. See the graph in the section that follows.
Apply to entire sound If your selection in the sound defines the noise sample, check this option. The effector changes the range of the selection so that it covers the entire sound after it has finished noise analysis using your selection.

Removing outstanding frequencies

The use of 100% pure background noise is always preferable, but if your noise sample unavoidably contains some instrumental tones that should not be taken as noise components, you might want to turn on the "Exclude outstanding frequencies" option.

With the option enabled, the effector further shapes the estimated noise floor, folding down as many spike-looking outstanding components as it has found, so that the shape of the floor will be flatter.

Note that this option removes some or all of the tonal components of the noise from the floor all together, as the effector cannot see the difference.

Example of noise floor flattening
Disabling the option
Enabling the option

Expected incoming damages

By applying this effect, some new damages may be introduced to the sound. The damages that may be audible are not limited to but include the following:

  • The tone of the sound may become slightly darker
  • The stereo image may be slightly altered
  • Some remaining components of the noise may be audible in an unnatural way

To keep the quality well, just avoid too much noise attenuation.

Calling back the removed noise

If you want to know something about the removed noise, e.g. what sound information was taken away together with it, then do the following:

  1. Copy the resulting sound to the clipboard
  2. Call back the original sound via the Undo operation
  3. Select the menu item [Edit] > [Subtractive Mix] from the main menu

The steps above will bring back the removed noise in front of you. You can view the spectral information with the spectrum view, or hear the noise with headphones, maybe after boosting its amplitude somewhat.


Notes on the Reverb Effect

This audio effect adds realistic echoes to a sound, simulating the physical phenomenon of reverberation.

This is a kind of sound reflection simulator. Physically possible reflection paths in a virtual room are collected via a simulation and the result is used for echo rendering. All of the calculation presupposes that the input sound is mono. The input sound can be of two channels but the content has to be mono if you expect the generated echoes to be coherent between channels. The effect always produces a sound in stereo. Use headphones if you need to check the resulting sound seriously.

The virtual room consists of several upright walls, a floor, and a ceiling. Each of these parts is something like a simple pane or board that has a flat surface. Some room properties can be set to what you like. You can make it a hall by setting its size very large.

The graph in the dialog box shows the shape of the room's floor which can be either rectangular or pentagonal. The horizontal and vertical scale lines show how long one meter is in the view. There are two circles on the graph. One represents a sound source and the other a stereo microphone set. You can drag either circle with the mouse to change its position.

The shape of the room

The room is formed out of several flat surfaces. The room can be a rectangular box (cuboid) or something irregular.

A cuboid room consists of three pairs of parallel surfaces, having all the corners cubical. In a room of this type, echoes tend to aggregate periodically and reverberations are perceived relatively sparse.

The room can be made irregular, given a pentagonal floor and ceiling. Any of the upright walls does not face in parallel with any of the others. This surface arrangement creates some degree of irregularity, and because of this irregularity, echoes each interfere with the others in a way more complex than they do in a cuboid room, and because of this increased complexity, echoes weaken their tendency toward periodic aggregation and reverberations become perceived denser.

The room can have a sloping ceiling. A pentagonal room with a sloping ceiling is fully irregular in the sense that every forming surface has no parallel relationship with any of the others. The use of a sloping ceiling makes echoes interfere in a further complex way compared to a level ceiling. The full use of irregularity makes a room best mimic a typical concert hall's way of reverberation.

If you choose a fully irregular room, it may take a very long time to calculate a set of reflection paths. For that reason, you might want the ceiling to be flat, which makes the calculation much faster at the expense of reducing the irregularity of the room.

Other settings

Some of the effect settings might need to be explained.

GroupSettingDescription
Room Echo Density If a group of ceasing echoes suddenly terminates, not fading away smoothly, it is probably because the value of this setting is not high enough. Note that the higher this value is, the longer it takes you to calculate a set of reflection paths. The calculation can take you very long when the room is fully irregular. This value defines the size of the smallest echo that is rendered.
Ceiling Slope Choosing a pentagnal floor and a sloping ceiling makes the room fully irregular. Making the ceiling too steep might give the impression of too dense reverberation.
Microphone Set HRTF-like Fx Enabling this option makes the direction cues of echoes emphasized somewhat. This option presupposes the use of headphones when listening.

It is difficult to predict the fluctuating range of the output waveform. To avoid waveform clippings, you might want to use float samples for temporary files. See How to Set the Sample Type for Temporary Files.

The effector takes advantage of multi-core CPUs. See How to Fully Employ Your Multi-core CPU. See also Performing Convolutions on GPU for GPU-accelerated convolution.

The execution process

The execution process of this reverb consists of three stages.

At the first stage, it calculates the set of reflection paths that matches the settings the user has made. At the next stage, it builds a set of room impulse responses, one for a channel, based on the result of the previous stage and the echo coloration settings. At the final stage, the convolution engine processes the input sound using the created impulse responses as FIR filters, generating the final output sound.

See the Technical Information page for more information.

Saving room impulse responses for later use

If you like, you can have the app write the room impulse responses to a file for later use. Note that it is the effect 'FIR Filter' that you use when you want to convolve a sound with it.

Because it is stored in a WAV file, it can be opened just like an ordinary sound file. To examine room impulse responses visually, the spectrogram view might be useful rather than the spectrum view.

Last but not least, please do not distribute any of the room impulse responses you have saved. The author considers them copyrighted materials which should be treated together with this app. Instead, you can introduce this app to your friends.


Performing Convolutions on GPU

If you have equipped your computer with a graphics processing unit (GPU), the app can utilize it to accelerate convolution operation, or sound filtering process, with long FIR filters. Your GPU may be able to accelerate the following two effects:

  • Reverb
  • FIR Filter

This GPU utilization option requires DirectX 12 being available. This means that you can enable this option only on Windows 10/11.

To enable this option,

  1. Select [Tools] > [Settings] from the main menu
  2. Click [Advanced] if the page is not in the foreground
  3. Check the item "GPU convolution in single precision"

In general, GPUs are primarily for game acceleration and do not perform well with double precision numbers (64-bit floating point numbers). Because of this, the app will not use any GPU for convolutions in double precision.

If you have equipped your computer with a GPU that is located outside the CPU, the app will use that external GPU in preference to the one integrated with the CPU.

CPU convolution vs. GPU convolution

Note that the app can utilize multiple CPU cores to accelerate convolution and convolution in that way might be able to outperform GPU convolution depending on the performance balance between them.

For CPU convolution and some other time-consuming tasks performed on CPU, you can set how many cores are maximally used. Because it is not cores but threads that applications can manage directly, the value you set is the number of threads to use. But each core is driven by a different thread, it also defines the number of cores to use. To set this value:

  1. Select [Tools] > [Settings] from the main menu
  2. Click [Advanced] if the page is not in the foreground
  3. Select the appropriate number from the list "Max. Number of Parallelism Threads"

If you are using a CPU, e.g. an Intel with hyper-threading cores, where the number of threads allowed to run simultaneously is greater than the number of physical cores, it is possible to set a value that exceeds the number of physical cores. But it is unknown whether or not such a setting improves performance.

Benchmarking the GPU

The button to run the benchmarking program is found on the Advanced page of the Settings dialog box.

The program simply repeats a particular convolution operation again and again for one second. The program does it twice, once on the CPU and once on the GPU so that you can judge which is faster. Then the program shows you how many sound segments the program was able to process on each computing device. The greater the count, the faster it is. You might want to run the program multiple times as the result can slightly vary.

The program performs convolution with a 2-channel FIR filter, the length of which is 221,184 samples long, or about 5 seconds long at 44.1 kHz. Both filter and sound are divided into segments, each of which is 4,096 samples long. Each FFT takes 8,192 samples, half of which is padding required for FFT convolution.


Enabling Dithering and Noise Shaping

When the app produces a 16-bit sound, downgrading a 24-bit or float sound, the app can optionally perform dithering on the sound to prevent a certain kind of distortion from being caused by reducing sample bits.

Dithering is done at the expense of making the signal-to-noise ratio (S/N ratio) of the sound slightly worse. In addition to dithering, the app can optionally use the noise shaping technique to reduce the quality loss introduced by dithering.

To enable dithering,

  1. Select [Tools] > [Settings] from the main menu
  2. Click [Quality] if the page is not in the foreground
  3. Check the item "Enable 16-bit dithering (not for temporary files)"

In addition, to enable noise shaping,

  1. Check the item "Enable noise shaping to improve S/N ratio"
  2. Mark the item "Shelving" if you have no idea of which is the best

Note that for temporary files, neither dithering nor noise shaping is performed regardless of your settings. You can use 24-bit or float samples for temporary files instead.

Dithering

To reduce the data size of a sound, we can re-quantize the sound with fewer sample bits. Re-quantizing a sound with fewer sample bits makes the sound less precise, or makes additional quantization noise come in. This incoming noise can have undesirable harmonic tones.

Dither is a small amount of random noise and can be added to the sound before re-quantizing. The dithering process randomizes quantization noise and prevents it from having any harmonic tones. Quantization noise in 16-bit samples is hardly audible thus the effect of dithering is not really obvious, but sensitive contents may benefit from dithering.

Noise shaping

The noise shaping technique can be used together to make both dither and quantization noise less noticeable. Summing up dither and quantization noise makes one single stream of noise. Once summed up, they always take a randomized form, becoming white in spectral color. The noise shaping process modifies the spectral shape of this summed-up noise so that it loses energy for the frequencies to which the human ear is sensitive and gains energy for the others.

For 44.1-kHz or 48-kHz sounds:

The app uses uniform-distribution dither, the fluctuation span of which is 1 bit. In principle, this dither becomes unsteady if the sound is nearly silent. But this dither is barely steady with a sine wave 120 dB smaller than a normalized sine wave, so in practice, this dither would work well with any music.

The shelving-type noise shaping filter attenuates the frequencies in the lower band by around 6 dB. Improving the signal-to-noise ratio by 6 dB equals improving the sample resolution by 1 bit, so sounds will be of 17-bit quality in effect with this filter.

For 96-kHz sounds:

The app uses triangular-distribution dither, the fluctuation span of which is 2 bits. This dither is always steady.

The shelving-type noise shaping filter attenuates the frequencies in the lower band by 17 dB.

The noise shaping curves of the other two types roughly trace the equal-loudness curve. The "3.5 kHz centered" type resembles the equal-loudness curve more closely than the other. The "6.5 kHz centered" type is roughly the inverse of the ITU-R 468 noise weighting curve. These types of filters lower the perceived noise level more than the shelving-type filter does.


About 'Project'

In the context of this app, a 'project' is one single set of sound files. A project can be saved and reopened later. The app can open and manage only one project at a time but can change the open project for another runtime.

By default, the app does not do any management that is project specific. You have to enable the project saving option manually on the [Basic] page of the Setting dialog box in order for projects to be saved. You can invoke the Setting dialog box by selecting [Tools] > [Settings] from the main menu.

When the option has been enabled, the app does not delete the temporary files on the exit. On the next start-up, it reopens all necessary files including the undeleted temporary ones and reconstructs the project. Likewise, the app does the same when it changes the open project for another runtime.

Each project has its own dedicated folder. The name of a project is the name of that dedicated folder. A new project can be created by entering the path name of a new folder into the edit box on the [Basic] page of the Settings dialog box. It is possible to open an existing project in the same way, but it would be more convenient to open an existing project by selecting its name from the project list [Tools] > [Project Folders] located on the main menu.

Note that when the user opens uncompressed sound files, the app does not make temporary files for backup. This means that if you move or delete any sound file that is necessary for some project, you will encounter a reconstruction failure when you open that project later. Since it could happen that you forget which project needs which sound file, you should not expect that you can keep a project for a long period without any trouble.


How to Set the Sample Type for Temporary Files

You can choose the sample type used for temporary files the app creates to store sound samples. It can be 16-bit integer, 24-bit integer, or 32-bit floating point number (float).

If the disk has a large amount of free space, the recommended choice is float because the use of floats can avoid waveform clippings.

To set the sample type,

  1. Select [Tools] > [Settings] from the main menu
  2. Click [Quality] if the page is not in the foreground
  3. Make a choice among the options on the list "Sample Type for Temporary Files"

How to Open 24-bit WMA files Using 24-bit Temporary Files

The professional version of the WMA encoder can attach the '24-bit' attribute to WMA files it produces. The app can optionally use 24-bit temporary files when opening such WMA files. To enable this feature,

  1. Select [Tools] > [Settings] from the main menu
  2. Click [Quality] if the page is not in the foreground
  3. Check the item "Import 24-bit WMAs preserving 24-bit quality"

How to Fully Employ Your Multi-core CPU

To improve speed, the app takes advantage of multi-core CPUs for some internal audio signal processing tasks which require much calculation. The app uses a team of threads to perform such tasks.

Applications are allowed to create and manage threads but not allowed to touch CPU cores directly. So an application creates as many threads as it needs and lets the OS utilize multiple cores to run that set of threads. For all of them to run in parallel, each of them needs to run on a different core. The CPU is maximally busy if the number of threads to run is equal to the number of cores.

To specify the maximum number of threads for such tasks,

  1. Select [Tools] > [Settings] from the main menu
  2. Click [Advanced] if the page is not in the foreground
  3. Make a choice among the numbers on the list "Max. Number of Parallelism Threads"

If you are using a CPU, e.g. an Intel with hyper-threading cores, where the number of threads allowed to run simultaneously is greater than the number of physical cores, you are allowed to set a value that is greater than the number of physical cores. But it is unknown whether or not such a setting improves performance.

The tasks for which the app uses this way of processing include:

  • Convolution with long filters like room impulse responses
  • Some internal tasks of the Reverb effect
  • Some internal tasks of the Click Removal effect
  • DSD 1-bit audio import and export

How to Employ the LAME MP3 Encoder

The app can optionally employ the LAME encoder when exporting MP3 files.

You need the DLL module called 'lame_enc.dll'. If you are using the 64-bit version of the app, you need the 64-bit version of that DLL.

You have to copy the DLL into the folder in which this app (nami.exe) is located. If you have chosen the default installation folder, its path should be the following:

C:\Program Files\Sandango\Sazanami

To make the app aware of the DLL,

  1. Select [Tools] > [Settings] from the main menu
  2. Click [Advanced] if the page is not in the foreground
  3. Check the item "Use LAME (lame_enc.dll) for MP3 export"

The next time you invoke the MP3 export dialog box, the app will present the LAME alternative.


How to Open MP4 Files

You can open MP4 files with this app. Note that the app takes any video file as read-only.

The app handles video files via DirectShow or Media Foundation. They are part of the OS. The Media Foundation MPEG-4 support is only available on Windows 7 or later.

To enable Media Foundation video handling:

  1. Select [Tools] > [Settings] from the main menu
  2. Click [Import] if the page is not in the foreground
  3. Check the item "Use Media Foundation for M4A/MP4 (Win7)"

If you prefer DirectShow, you have to install two kinds of DirectShow filters:

  • A filter that works as an MPEG-4 decoder (that supports H.264 and AAC)
  • A filter that works as an MP4 splitter

Some freeware filters should be available on the net. Note that a DirectShow filter is a DLL module and if you are using the 64-bit version of the app, you have to install the 64-bit version of DirectShow filters.

Installing an FLV splitter in addition makes it possible for the app to handle FLV files that contain MPEG-4 videos.


Help Document Translation

The Japanese version of the Help document is also available online:

https://ss1.xrea.com/sandango.s322.xrea.com/namihelp/

If you like, you can use machine translation services on the net to translate the document. If the translation by Microsoft's engine does not come up to your expectation in quality, try Google's.


Last Update 2026/04/01

[Sazanami Help]

ReadMe
Usage Notes
Settings
Supported File Formats
Technical Information
Version History
Donations Welcome
Help in Japanese

Copyright (c) 2026 Narumi Watanabe.
All Rights Reserved.