For both reading and writing:

• Formats: WAVE (*.wav), AIFF/AIFC (*.aiff/*.aifc) and AU Sun/NeXT (*.au).
• Encodings: 16-bit (signed integer, PCM), 24-bit (signed integer, PCM) and 32-bit float* (PCM)
  • *For 32-bit, only IEEE float encoding with sample values normalized to the -1.0 - 1.0 range is supported (known as type 3 for WAVE files); other 32-bit encodings (32-bit signed integer, or float in the +/- 32768 range) are not supported.
  • Compressed encodings within WAVE, AIFF/AIFC or AU files are not supported.
• Audio files of more than 4 GB are supported in the WAVE format, using the "RF64" extension, and in the AU format.

For reading only: MP3, FLAC, OGG, OPUS, M4A (AAC and ALAC). Proton might not recognize multi-channel or high sample rate audio files in these formats.

Warning: the AU file format used by Proton (also known as Sun/NeXT audio, SND, or "basic audio") is unrelated and not compatible with the *.au files used internally by the Audacity free editor, although they use the same file extension. Audacity can both load and export audio files in the AU format used by Proton though.

Yes. Proton supports mono, stereo and multi-channel files (including 5.1 and 7.1) as long as the individual channels correspond to stereo/spatial information (and not to different "tracks"), and as long as there is enough memory.

If the channels correspond to different "tracks" (such as vocal, bass, etc), it is preferable to split the tracks into individual files and to transform them separately. The reason is that Proton locks the phase of all channels of a multi-channel file in order to preserve the stereo (or spatial) field. Locking the phase between unrelated tracks on the other hand might result in a deterioration of the quality, which is why unrelated tracks should be transformed individually - using separate files. If a given track is stereo, it should be extracted into a stereo file rather than in two mono files.

Warnings:

• Proton does not preserve meta-data related to channel/speaker configurations of multi-channel audio files.
• Surround formats that "pack" more than two virtual channels into a stereo file (such as Dolby Surround or Dolby Prologic) might work but this has not been tested.

• Sample rate: from 32 kHz to 192 kHz.
• Bit depth: 16-bit (signed PCM), 24-bit (signed PCM) and 32-bit IEEE float (with samples between -1.0 and 1.0 - type 3 for the WAVE format). The 32-bit float format is always used internally.

The maximum is 8 times slower or 8 times faster.

Note that within the application, the slider (see (1) in the figure below) only allows you to choose from 2 times slower to 2 times faster (which is the recommended range). You must type a value directly in the number field (2) next to the slider to choose a higher ratio up to the maximum. Transforming more than 2 times slower or faster might result in lower quality.

There is no limitation as long as you have enough disk space and enough processing time. Proton has been tested with audio files of several hours of duration, and might be able to process even longer files.

There are a few caveats though:

• The AIFF/AIFC (*.aiff, *.aifc) file formats do not allow files of more than 2 GB. You must save them in the WAVE (*.wav) or AU (*.au) format. The application will ask you to change to the WAVE (*.wav) format if an AIFF/AIFC file would get larger than 2 GB.
  • See Section 3.2.3 to change the output file format before starting the transformation.
• Some applications may not support reading WAVE files of more than 4 GB or AU files of more than 2 GB. For WAVE files of more than 4 GB, Proton automatically uses the "RF64" extended format.
• You must have enough disk space for the sum of the sizes of:
  • the input file,
  • the temporary output file (always in 32-bit float raw format) - see Section 4.3.2 to change the location of temporary output files,
  • the output file in the specified target format (can be 16, 24 or 32-bit) - see Section 3.2.3 to change the location of output files.
• The Mini Player (see Section 4.1) can be slow when seeking into input files of long duration. You may want to use your favorite audio player instead.

No. However, you can save in a 24-bit or 32-bit format (see Section 3.2.3 ) and apply dithering afterwards using your favorite DAW.

The length of the transformed file might drift from the exact theoretical length by a few samples due to internal buffering. However the actual time stretching ratio is accurate to a single sample event with audio files of one hour of duration.

Even if the time stretching ratio is displayed as rounded in some parts of the user interface, the exact entered value is remembered and used, up to the 10^th decimal point.

This depends on many factors. Using a 44.1 kHz stereo file, it generally takes about 4 seconds to produce 1 minute of audio on a 3 Ghz processor using the default quality setting, meaning about 15 times faster than real-time.

In comparison to the default settings (Quality, Poly, transient detection enabled; on a 44.1 kHz stereo file):

• The Fast quality setting is about 80% (1.8 times) faster than the Quality setting.
• The Solo mode is about 15% faster than the Poly mode.
• Disabling transient detection is about 25% faster. Note than changing the transient sensitivity has no effect.
• The Smooth mode is the slowest, about 4 times slower than the Poly mode. It might not achieve real-time on old computers.
• Transforming audio files with more (or less) channels takes proportionally more (respectively less) time, approximately.
• Transforming audio files with a higher (or lower) sampling rate takes proportionally more (respectively less) time.
• Transforming a 16-bit, 24-bit or 32-bit float audio file takes the same time (bit depth, or resolution, has no impact on processing speed).
• On a multi-core system, multiple files can be transformed simultaneously, saving some time. See section 4.3.2 to change multi-core settings (number of processing threads).

Yes, except the Smooth mode on slow computers.

Note

By real-time processing, we mean that time stretching a given duration of audio takes less time than that duration. For example, time stretching 10 seconds of audio takes less than 10 seconds. This basically means that you can play the result immediately and without interruption, as it is being transformed.

Another definition of real-time is to transform audio and play the result on the fly as it comes from the microphone or audio input. This is not possible with Proton, and does not make a lot of sense for time stretching anyway.

No, the current version is only available as a stand-alone application. A future version might be available as a plugin.

Some DAW might support drag and dropping audio files or tracks to the "Input Files" list of Proton, and then drag and dropping back the transformed files or tracks from the "Output Files" list. While not as practical as a plugin, this feature might simplify the workflow when working with both a DAW and Proton.

Yes and no.

• Proton uses only a single core (at a time) when processing a single audio file.
• However, when transforming multiple files, Proton can use multiple CPU cores simultaneously to process different files simultaneously.

For example, if there are 4 CPU cores, and you transform 10 files, Proton will start transforming the 4 first files at the same time using the 4 available CPU cores, and the 6 other files are held in a waiting queue. As soon as one of the files is finished, the corresponding CPU core immediately starts processing the next file in the queue, keeping all the CPU cores busy.

Note that by default, Proton uses at most half the number of logical CPU cores available, which might not always correspond to the number of physical CPU cores available. This setting can be changed in the global settings (see section 4.3.2)

Proton does not implement pitch shifting. However if your DAW provides resampling, it is possible to do pitch shifting by combining resampling with time stretching. For best results, you should always resample first, and then time-stretch; regardless of whether you want to pitch shift up or down. Please note that pitch shifting that way will not preserve the formants.

Whatever XY is, the answer is probably no.

Proton does one thing, and does it well. That one thing is time stretching audio files.

• Proton only does time stretching. It does not do pitch shifting.
• Proton does not support sliding time stretching (time stretching by an amount that is varying with time).
• Proton is a standalone application. It is not available as a plugin.
• Proton can only write audio files in the WAVE (*.wav), AIFF (*.aiff, *.aifc) and AU Sun/NeXT (*.au) formats, with PCM encoding (including 32-bit float). It can read these formats and a few compressed formats: MP3, FLAC, OGG, OPUS and M4A.
• The Mini Player included in Proton might not be able to play some multi-channel or high-resolution audio files. But they can be transformed without problem.

Proton can be downloaded:

• From the Proton home page.
• Coming soon (once it is no longer a beta version): KVR Marketplace, Microsoft Store, Mac App Store.

• Operating system:
  • Windows 64-bit: Windows 11 or later, Windows 10, Windows 8, Windows 7 (SP 1).
    • The version from the Microsoft Store is only available for Windows 10 or later.
  • Mac OS: Mac OS X 10.11 "El Capitan" or later (64-bit Intel), or macOS 11 "Big Sur" or later (Apple M1).
  • Note: Other versions are available, but they have been less extensively tested: Windows 32-bit, Linux (depends on a JRE version 11 or greater), Generic (requires a JRE version 8 or greater). Use these at your own risk.
• Memory:
  • Minimum: 50 MB available for Proton. This allows transforming four stereo, 48 kHZ audio files at a time (on a computer with at least 4 CPU cores).
  • Recommended: 200 MB.
• CPU speed: Any, but at least 1 GHz is recommended. The faster the CPU, the less time it takes to transform a file.
• CPU cores: The more cores available, the more files Proton can process at the same time. Note however that multiple cores will not make the transformation of a single file faster. Proton can only use multiple cores at the same time to transform multiple files at the same time.
• Disk: Disk speed is generally not important. Disk capacity on the other hand must be sufficient to store the original, transformed and temporary (being processed) files.

Proton only requires an internet connection when you ask it to:

• Check for updates
• Open the Home Page
• Display the online documentation

Apart from that, no internet connection is required while using Proton.

Proton is available for free.

Different parts of Proton are licenced under different licences:

• The binary and the installer are licenced under the CC-BY-ND licence.
  • You can redistribute the binary or the installer (CC), but you are not allowed to alter them (ND).
  • You must give appropriate credits and link to the original licence (BY).
• The time-stretching algorithm is open source, licenced under the Eclipse Public License.
  • The source code of the time-stretching algorithm can be downloaded here.
• The source code of other parts (except third-party libraries with their own licences) is Copyright 2024 Nicolas Juillerat, all rights reserved.
  • This basically means "proprietary".

While the process of time stretching mostly preserves the loudness, it frequently increases the level of individual peaks, which would result in clipping. This is especially true with modern mastering that have been brick-wall limited with virtually no headroom. By default, the Proton application automatically reduces the loudness of the transformed files as necessary to prevent clipping (peak normalization).

This behavior can be modified by disabling the "Prevent Clipping" option in the transformation settings. See section 3.2.3 to change the setting on individual files, or section 4.3.1 to change the setting globally. Disabling this option might result in clipping on the transformed file, unless the output file format is set to one of the 32-bit float formats. Note that with these formats, samples might still be above the ±1 maximum "normalized" value and do clip when played, but the true values are not lost and a DAW can then normalize them properly or apply a suitable dynamic range compression.

In general, we recommend you transform audio files with sufficient headroom to prevent clipping or automatic lowering of the volume. Audio files normalized to EBU R128 for instance rarely clip when transformed. Dynamic range compression, brick-wall limiting and any other form of loudness maximization is better done after time stretching, if ever.

Make sure Proton is not saving files in PCM 32-bit floating-point format. This format is recognized by most professional DAW, but might not be recognized by consumer audio players.

You may want to save the transformed files in PCM 16-bit or 24-bit instead. See section 3.2.3 to change the output format in which transformed files are saved.

If a file is currently being transformed, select it in the "Output Files" list, and click on the Info... button. In the "Transformation" tab, there is an item labeled "Processing time" that displays both the time elapsed so far on this file, and an estimation of the time remaining (ETR). The estimation might be inaccurate or unknown at the beginning of the processing, but it gets more accurate as the processing advances.

If multiple files are processed, there is no estimation of the total time remaining.

• Microsoft Store version: by default, the transformed files are stored in the Music/Proton folder in your home.
• KVR version: by default, each transformed audio file is stored in the same directory as the original, untransformed file.

It is possible to instead explicitly specify an output directory in which the fully transformed files are stored. See section 3.2.3 (individual file setting) and 4.3.1 (global setting). See section 4.3.2 (global setting only) to change where the partially transformed files are stored.

Notes:

• Right-clicking on a transformed file (in the "Output Files" list) and choosing the Open containing folder menu will open the containing folder in the file explorer.
• Moving the mouse over a transformed file (in the "Output Files" list) and waiting one second will show the full file path on a tooltip

On the drop-down under "Time stretching factor", choose Video FPS from/to (FPS = [video] Frames Per Second). Enter the original video frame rate in the "From:" field and the final video frame rate in the "To:" field.

For example, a typical NTSC to PAL conversion plays a video track originally at 23.976* FPS (NTSC) on a 25 FPS system (PAL). In that case, to stretch the audio accordingly, enter 23.976 in the "From:" field, and 25 in the "To:" field.

*This value originates from a 29.97 FPS (the true NTSC FPS) video track converted by an approximate 24/30 ratio. An exact conversion of a 29.97 to a 25 FPS video track is generally difficult to perform with good image quality, which is why the resulting video track is typically at 23.976 FPS instead of 25. However, in case the video track is converted exactly (with an exact 29.97/25 ratio), there is no need to stretch the audio track.

When in doubt, you can also choose Length from/to in the drop-down under "Time stretching Factor", and enter the exact initial and final length of the video to properly stretch the audio accordingly.

Note: Proton does not handle video or movie files. You must extract the audio tracks from video files with your video editor in order to stretch them using Proton.

Yes and no. There is no "preview" feature.

However, when you hit the Transform button to transform a file, you do not need to wait until the file is fully transformed to play the result.

Just select the file (in the "Output Files" list) and hit the Play button (under the "Output Files" list) and Proton will play as much audio as is already transformed.

The player itself shows both the playback position and the amount of audio produced yet (the "processing" position); see section 4.1. The current duration of the transformed audio is also displayed next to the file name in the "Output Files" list, and is regularly updated until the transformation is finished.

If the file is not fully transformed but you are not satisfied with the result, right-click on the file in the "Output Files" list, and choose Abort Processing from the contextual menu to cancel the transformation.

If you want a quick and low-quality preview of the whole transformed file, see section 3.2.2 to lower the processing quality (which results in faster processing).

Choose the File, Settings... menu, select the Processing Tab; under Prevent system sleeping, be sure the ☑ While tranforming files option is checked. Else, the system might go to sleep before all files have been transformed by Proton, which will pause the processing until the computer is awakened again. If this option is not visible, you may need to update the application.

Also note that by default, Proton transforms files using low priority (or background) threads, to make sure the system does not get unresponsive. The consequence is that other CPU-intensive processes that do not run with background priority might sometimes "steal" CPU cycles from Proton. You can also change the priority of the processing threads used by Proton in the global settings described in section 4.3.2.

Note: because of a bug, sometimes the GUI does not respond when returning from sleep mode or lock screen. Just move the Proton's main window and the problem should go away.

Generally, because the sample rate or the number of channels is not supported by the Mini Player integrated in Proton.

This can also happen if another application has exclusive access to the default sound card.

The Mini Player in Proton is a basic player: it plays through the operating system's default audio output (using alsa) , and does not try to detect or use advanced sound cards automatically. If the default audio output does not support the sample rate or number of channels, the player will give an error. Note that Proton will still be able to time-stretch the files even if it cannot play them.

In Proton, right-click on a file, and choose the Open containing folder menu to show the file in the file explorer. You can then play it with your favorite player or DAW.

• The default settings, Quality and Poly gives the best results for polyphonic musics.
• The Solo mode usually gives better result on solo tracks, but can also give better results on polyphonic musics if the time-stretching ratio is close to 1. For speech, the Fast setting sometimes gives better results.
  • See section 3.2.2 to change the transformation setting on individual files.
  • See section 4.3.1 to change the default transformation setting for all new files.
• If the transients are softened (or exagerated), you may want to adjust the transient sensitivity setting too.
• If that is still not enough, consider using on of our commercial time-stretching softwares, Quantum Audio Time Stretch (Quantum) or Tachyon Audio Time Stretch (Tachyon).
  • Quantum is an application similar to Proton, and usually gives higher quality results. It is not free though, but a demo version is available. It has 5 quality settings, most of them are slower than Proton.
  • Tachyon is our reference time-stretching application, optimized for the best possible quality. Note however that Tachyon is much slower than Proton, and is more expensive than Quantum.
  • A Tachyon licence can be used to unlock Quantum (but not the reverse). Hence you can either buy Quantum only, or both Quantum and Tachyon.

• Check that your are running on AC power and not on battery.
• Changing to Fast and/or Solo, or disabling transient detection can all make the processing faster, although with a lower quality.
  • See section 3.2.2 to change the quality setting on individual files.
  • See section 4.3.1 to change the default quality setting for all new files.

Choose the File, Settings menu; select the "Appearance" tab; Select Flat Dark in the "Appearance" drop-down, and click on Save. Now Proton looks professional.

Proton was not developed by a big company, but by a single person (Nicolas Juillerat, from Fribourg, Switzerland) working on it about one day a week. As such support is not expected to be the same as that of big companies.

You may get answers by:

• Reading this user manual. Click here to go to the table of content.
• Ask your question in the KVR Forums.
• Ask your question by email to tats(at)tachyon-sonics(dot)com. Mention "Proton" in the subject. Answer may take 1 to 7 days.