AI Mastering Blog

We compared LANDR and AI Mastering in sound quality.

Overview

We proposed an index that can objectively evaluate the mix MEI 20190207.

We compared the sounds mastering with AI Mastering and LANDR at MEI 20190207.

We found that AI Mastering has higher MEI 20190207 than LANDR.

AI Mastering has a tendency that the loudness range is larger than the LANDR, the Boominess is small, the Depth is small, and the Warmth is small.

* Since there are comparative sounds in the other people, please listen

Comparison method

Mastering various sounds with LANDR and AI Mastering and comparing the results with various indicators.

Sound to be compared

We chose the sound to be compared from the following mix evaluation data set. This mix evaluation data set includes multiple mixed audio for various songs and subjective evaluation results by multiple people for each mixed audio.

In the mix audio license, CC BY's, we selected the one with the largest loudness range for each song and the one with the lowest average subjective rating as the comparison target tone.

The reason is that it is easy to master without artifacts when the loudness range is large, and there is a mismatch of automatic mastering when the subjective evaluation is low.

Please see the GitHub repository below for a specific mix list.

MixBrowser

THE MIX EVALUATION DATASET

index

MixEvaluationIndex20190207 (MEI20190207)

MixEvaluationIndex 20190207 (MEI 20190207) is an objective evaluation index of mixed audio constructed using subjective evaluation data of The Mix Evaluation Dataset. It is an evaluation index of mixed audio, but I think that it can also be used for evaluating mastering audio. It is intended for comprehensive evaluation. It is the main indicator in this comparison.

MEI 20190207 is calculated by the weighted sum of various indices. The original indices are the spread covariance matrix of the spectrum, the mean of the spectrum, Hardness, Dissonance. Simply put, I calculate it based on the shape of the spectrum, the dynamic range, the spread of space, the bandwidth of the attack, and the amount of distortion.

The mixed audio used for weight learning is all the mixed audio that is published in MixBrowser, with preview audio. Some preview audio was 404 Not Found.

MixBrowser

Loudness

It is the loudness defined by ITU-R BS.1770. Depending on the platform to be delivered and how the user listens, it is highly likely that songs with loudness are more likely to be played with louder sounds as compared to other songs. It sounds better as you play with loud sounds.

If the sound quality is the same, the loudness should be large.

Other indicators

Loudness range, True Peak

Mastering setting

Please see GitHub below.

Comparison result

Change amount of MEI 20190207

It is the average of MEI 20190207 change with respect to the original for all songs. AI Mastering tends to have higher MEI 20190207 than LANDR.

Loudness vs. Loudness range

I plotted the average of all songs in loudness and the average of all songs in loudness with a scatter plot. In general, the loudness and the loudness range are in a trade-off relationship, but the decrease in the loudness range when AI Mastering has a higher loudness than LANDR is small.

True Peak

True Peak (inter sample peak) is an average of all songs. If True Peak is larger than 0 dB, it may be distorted due to re-encoding etc, but there seem to be cases where both AI Mastering and LANDR exceed 0 dB. If you set Ceiling to True Peak in AI Mastering, you can prevent True Peak from exceeding 0 dB, so you can avoid degrading sound quality. LANDR is probably impossible to avoid because there is no such setting.

Change amount of Dissonance

Dissonance is an index to measure dissonance degree. It is used to calculate MEI 20190207, the lower the Dissonance, the higher the MEI 20190207.

If you set the mastering level to 1 in AI Mastering, Dissonance seems to increase. Setting the mastering level to 0.5 will result in an increase equivalent to LANDR.

Change amount of Hardness

Hardness is an index to measure hardness of sound. It is used to calculate MEI 20190207, and the higher the Hardness is, the higher the MEI 20190207 is. Both AI Mastering and LANDR seem to increase Hardness.

Changes in Boominess

Boominess is an index for Boomy degree. The contents are Booming Index proposed below. It is not used for calculation of MEI 20190207.

Booming index as a measurement for evaluation booming sensation

AI Mastering tends to lower Boominess.

Change amount of Brightness

Brightness is an indicator of brightness. It is calculated by the linear combination of the logarithm of the energy ratio of the high frequency component to the total energy and the logarithm of Spectral Centroid. It is not used for calculation of MEI 20190207.

D5.2: First prototype of timbral characterisation tool for semantically annotating non-musical content

AI Mastering and LANDR tend to raise Brightness.

Change amount of Depth

Depth is an indicator of depth. It is defined in D 5.2 below. According to D 5.2, the depth has spatial meaning and frequency characteristic meaning, but this Depth index represents only frequency characteristic meaning. It is not used for calculation of MEI 20190207.

D5.2: First prototype of timbral characterisation tool for semantically annotating non-musical content

According to the definition, Depth will increase if there are many low frequency components. AI Mastering tends to lower Depth.

Amount of change of Warmth

Warmth is an indicator of warmth. The following is an implementation. It is not used for calculation of MEI 20190207.

Timbral_Warmth.py (Github)

AI Mastering tends to lower Warmth.

Comparison of sound after mastering

For each song, I picked up three of the original, LANDR with the largest MEI 20190207, AI Mastering with MEI 20190207 the biggest one. Since loudness is not aligned, please be careful of the bias due to the difference in volume.

All sound lists are below. Please try MEI 20190207 whether the high sound is really good sound. The license notation of each song is described under the audio directory of Github.

ai-mastering / mastering_comparison (Github)

In The Meantime

Original

AI Mastering Best MEI20190207

LANDR Best MEI20190207

Lead Me

Original

AI Mastering Best MEI20190207

LANDR Best MEI20190207

Not Alone

Original

AI Mastering Best MEI20190207

LANDR Best MEI20190207

Pouring Room

Original

AI Mastering Best MEI20190207

LANDR Best MEI20190207

Red To Blue

Original

AI Mastering Best MEI20190207

LANDR Best MEI20190207

Github

Detailed information is listed below.

ai-mastering/mastering_comparison (Github)

Note

What is written as "AI Mastering" on the graph or Github represents AI Mastering.

Summary

I compared LANDR and AI Mastering.

We will introduce PEAQ (Perceptual Evaluation of Audio Quality).

PEAQ

Can PEAQ perceive the difference between the reference sound and the target sound? It is a method to objectively evaluate by calculation. I think that it can be used to evaluate the performance of the limiter and the performance of the mp3 encoder.

PEAQ (Wikipedia)

ITU-R BS.1387

PEAQ license

According to Wikipedia, PEAQ seems to be protected by patents. So you can not use it freely. According to the site below, you may be able to sign a license via OPTICOM.

PEAQ (OPTICOM)

Licensing (OPTICOM)

Implementation of PEAQ

In the following article, accuracy of multiple PEAQ and calculation time are compared. It seems that GstPEAQ is good for accuracy.

GstPEAQ – An Open Source Implementation of the PEAQ Algorithm

GstPEAQ

It is a PEAQ implementation created as a GStreamer plugin. It is implemented in C language and the license is LGPL 2.

HSU-ANT/gstpeaq (Github)

PEAQ test data

The zip file that can be downloaded from the link below contains test data. According to the license notation in the zip file, it seems that it should not be used other than evaluation of PEAQ implementation.

ITU-R BS.1387

Summary

We introduced about PEAQ. I thought about introducing it to AI Mastering, but I abandoned it because I can not use it freely.

I will introduce Audio Commons.

Audio Commons

"Audio Commons Initiative" is an initiative aimed at delivering audio content provided under the Creative Commons license to the creative industry. I do not know the meaning of Initiative, but I think that it is probably used in a sense like organization.

Audio Commons introduction summary

· Reusable audio contents (sound effects, field recordings, audio samples, songs) should be originally more, but there are few current situations.

· The cause is a lack of shared culture on content reuse and technical problems (search, license clarification).

· Audio Commons Initiative solves those problems and encourages reuse of audio content.

Audio Commons (official website)

Audio Commons (Twitter)

Audio Commons Ecosystem

Audio Commons Ecosystem (ACE) is an ecosystem of audio content, users and tools supported by Audio Commons Initiative. Audio Commons Ecosystem facilitates the reuse of audio content provided under the Creative Commons license in professional, non-professional and personal areas.

Character of ecosystem

Content Creator: a person who creates content. Composers etc.

Content Provider: the person who delivers the content. Jamendo, Freesound, etc.

Content User: The person who uses the content. Game producers etc.

Audio Commons ontology

Audio Commons ontology is an ontology of audio content used in Audio Commons Ecosystem. With Audio Commons ontology, you can annotate audio content (give metadata).

Participants in Audio Commons

In addition to the university, people belonging to Waves, Jamendo, AudioGaming are participating in Audio Commons.

Audio Commons Team

Audio Commons' Github repository

Audio Commons distributes various libraries with Github.

Audio Commons (GitHub)

An interesting tool of Audio Commons

Timbral Explorer

Multiple audio samples are visualized in two dimensions. A circle represents one audio sample and is arranged so that sounds resemble one another. Clicking the circle will sound. If you click a circle nearby, a similar sound will sound, and clicking on a circle in the distance will sound a sound that is not similar.

If you can use it from DAW, it seems easy to select audio samples.

Timbral Explorer

Summary

Audio Commons introduced.

We introduce speakers "Mirai speaker" for patients with hearing loss and similar products.

What is "Mirai Speaker"?

"Mirai speaker" is barrier-free speaker of sound. It is a product of Sound Fun Corporation.

Sound fan official website

How do I acquire 'Mirai Speaker' by yourself?

It is possible to rent it at 2,980 yen / month.

"Mirai speaker" rental

Principle of "Mirai speaker"

I think that it is based on the following patents. The details of the principle seems to be unknown.

Patent (Universal speaker)

What is Comuoon?

"Comuoon" is a dialogue support equipment. It is a product of Universal · Sound Design Co., Ltd.

Universal · Sound Design Corporation Official website

How can I get "Comuoon" by my own?

It seems that it can purchase at 7,600 yen + 36 * 5,800 yen = 216,400 yen. It seems there are times when you can use "daily living gear tool benefit system".

About purchase method (Comuoon official website)

Principle 1 of "Comuoon" - Conversion of sound

comuoon clears the sound input from the microphone clearly and converts it into a clear sound that is easy to hear.

Source

It seems that it converts the input sound so that it is easier to hear and output it. If you estimate from the images posted on the citation source, there is a possibility of boosting the important band for listening by EQ.

Principle of "Comuoon" 2 - Transverse Wave Speaker

"Susumushi speaker" is a speaker that applied the principle that Sumushi grinds wings and produces sounds, seems to be a hint of Comuoon.

Copernicus products for hearing impaired people to improve difficulty in hearing

I found a paper by Prof. Yoshihiro Muto, who introduced "Susumushi speaker" in the above mentioned article. In the thesis, the acoustic characteristics of the underlay bent to the music box is analyzed. It is similar to the structure of "Mirai speaker". Is the root of "Comuoon" and "Mirai speaker" the same?

"Transverse Wave Speaker" having a structure combining a piezoelectric element and a diaphragm is also introduced. Perhaps this is what is called "Sumushi speaker" or "Yokonowa speaker"?

Known and unknown phenomena of nonlinear behaviors in the power harvesting mat and the transverse wave speaker

It seems that the result of measuring the sound intensity for each distance is consistent with the result when assuming that transverse waves in the air are generated. However, since the method of deriving the three parameters used in the simulation is not described and the validity of the derived parameter is not discussed, it is not known whether the simulation result can actually exist or not.

There is also a theory that there is no transverse wave of sound, so it happens that the calculation result and the experiment result coincide, it is possible that the principle is actually based on another principle.

Speaker which can generate plane wave

Some speakers have speakers that can generate plane waves. As a result of examination, it is called by various names such as flat speaker, flat speaker, plane wave speaker, etc. There seems to be no unified way of calling.

Supposing that the essential point of "Mirai speaker" and "Transverse Wave Speaker" is plane waves rather than transverse waves, other speakers that can generate plane waves are likely to be effective for hearing loss patients as well.

I examined speakers that I can purchase individually.

F. P. S.

FPS online store

TOA Corporation

TOA PW-1230 DB Plane Wave Speaker, Black

Relationship between hearing loss and plane wave

If a plane wave comes out from "Mirai speaker" or "Transverse Wave Speaker", there is a possibility that the plane wave can be heard for the patient with hearing loss. If so, think about the reason.

Hypothesis 1 - Because it is difficult for the volume to attenuate with distance

Compared to spherical waves, plane waves tend to attenuate in volume depending on distance. Even if it is far from the speaker, it is easy to hear it, so there is a possibility that it may be easy for the hearing impaired patient to hear.

We can verify this hypothesis by comparing the easiness of listening with the plane wave and the spherical wave adjusted to the same volume at the position of the subject with respect to the subject in the same place.

Hypothesis 2 - Volume change when the ears are slightly moved is small

It is similar to hypothesis 1, but due to the influence of ear structure and room reflection, the plane wave is compared with the spherical wave, the volume change in the inner ear when slightly changing the direction of the ear or the ear position compared with the spherical wave There is a possibility that it is small.

There seems to be no one who lives with the ear position and direction fixed perfectly, so the magnitude of the volume change when moving them a little can affect the ease of hearing.

Experiment of hypothesis 1 can be verified if it is done with perfectly fixed condition of ear position and direction and condition moving little by little.

Hypothesis 3 - Because there is little reverberation

Plane waves have strong directivity, so there are few reflections in the room and reverberation sound may be reduced. There is a possibility that the amount of reverberation influences the ease of hearing.

It can be verified by measuring the amount of reverberation of the spherical wave and the plane wave and comparing the hearing easiness with the sound corrected so that the reverberation amount becomes the same with the reverb and so on.

Hypothesis 4 - Because the reverberation is simple

It is similar to hypothesis 3, but there is a possibility that the structure of the reverberation influences the ease of hearing rather than the amount of reverberation. For example, with reverb and short delay, the ease of hearing may be different even with the same amount of reverberation.

It is possible to verify by measuring the impulse response of the spherical wave and the plane wave and comparing the hearing easiness with the sound that simulates the structure of the reverberation by convolution and so on.

Summary

We introduced a speaker "Mirai speaker" for patients with hearing loss and similar products.

※ Price etc is information at the time of article writing. For the latest information please visit the official website.

We will introduce SoundBridge, a free DAW released in 2018.

What is SoundBridge?

SoundBridge is a free DAW released in 2018. Compatible with Windows and Mac. According to this , SoundBridge seems to have been made over three years.

SoundBridge official website

Features of SoundBridge

Because it monitors with music teaching materials, DAW itself is characterized by being able to use it for free .

How to use SoundBridge

Free member registration for SoundBridge

We will register as a free member on SoundBridge official website .

Download SoundBridge

When the member registration is completed, the download link of the SoundBridge installer according to the OS used is displayed. If the OS automatic judgment is wrong, you can download the SoundBridge installer for each OS from the My Accounts screen.

Installing SoundBridge

Launch the SoundBridge installer and install it according to the screen. All default settings are OK.

Starting SoundBridge

When you start SoundBridge, the following screen will appear, so enter your account information and log in.

SoundBridge sample project

When installing with the default setting, the sample project is installed in the Documents directory. Open this from the File menu in the lower left corner of SoundBridge. Some projects could not be opened due to errors. ExampleProject.soundbridge opened.

I will play it.

If you try to edit the sample project popup prompting you to register in SkyTracks.io is displayed and you can not edit it. Please open a new project if you want to edit it.

Piano roll of SoundBridge

It is the same as a general DAW piano roll.

SoundBridge Mixer

It is the same as the general DAW Mixer.

What is SoundBridge Academy?

SoundBridge Academy is a paid online video material that you can learn about music production. It may be useful to attach the DAW project file. It might be possible to see a view called DAW with a rich paid manual.

SoundBridge Academy

SoundBridge's reputation

I had issues with Ozone 8 plugins freezing up the application v1.09, and a simple email generated a quick response and a timely update to fix. 5 stars for customer service on that.

Source

It seems that if you report a bug it will be fixed by haste.

I'd happily pay a monthly fee to have access to your learning content, tutorials and you can keep it interesting for novice learners with monthly perks / sample kits etc.

Source

The place to learn music theory seems to be the charm of SoundBridge.

Does SoundBridge charge for future?

In an effort to make all of this accessible to as many many people as possible, we have decided to make the software completely free.

Source

Although there is no guarantee that it will not be charged in the future, since it monitors with music teaching materials, if that method goes well, DAW itself is likely to be free for a long time.

Ableton and PreSonus offer Tips and helpful content on free blogs and monitize with DAW. SoundBridge may be able to view it as a reverse.

Summary

We introduced the free DAW, SoundBridge released in 2018.

We will introduce movies and demonstration audio that understand the effect of "ClearMixer".

"ClearMixer" introduction video

"ClearMixer" introduction video commentary

Status setting

In the movie, Band Noise (narrowband noise) and Sine Wave (sine wave) are ringing at the same time. Usually in this situation, you can hear turbidity when the volume of Band Noise and Sine Wave are about the same, but demonstrate that using "ClearMixer" reduces turbidity.

Control of interference by "Intensity" parameter

From 0: 21 to 0: 37, I manipulate the "Intensity" parameter of "ClearMixer" and control interference between Band Noise and Sine Wave. Increasing the "Intensity" of "ClearMixer" decreases the volume of Band Noise in the band where Sine Wave is ringing. I think that it is easy to understand when listening with headphones. You can see it with the spectrum analyzer of "ClearMixer" shown in the movie. This makes Sine Wave easier to hear.

Switching output sound by "Output" parameter

From 0: 38 to 0: 47, the output sound is switched by operating the "Output" parameter of "ClearMixer". Sounds that Dry does not process with ClearMixer, and Wet represents sounds processed by ClearMixer. Wet - Dry represents the difference sound before and after ClearMixer processing. Interference represents the interference component. I think that interference reduction effect is easy to understand when listening to Wet - Dry.

Listen only to Band Noise

From 1: 00 to 1: 12, I mute Sine Wave and play back only Band Noise, then I control interference. I think that the interference reduction effect by ClearMixer is more clearly understood by Sine Wave being muted.

Change the volume of Sine Wave

1: 18 ~ 1: 38, the volume of Sine Wave is changing. Looking at the spectrum analyzer of "ClearMixer", you can see that the interference reduction amount of Band Noise changes depending on the volume of Sine Wave. I think that I feel almost uncomfortable because it changes smoothly.

"ClearMixer" demo audio

Original audio

Audio after applying "ClearMixer"

Audio (over-applied) after applying "ClearMixer"

Commentary

Although the difference between the original audio and the audio after applying "ClearMixer" is a subtle difference, I think that it is easy to hear by paying attention to the interference between the base overtone and the midrange of the piano. I think that it is clean as a whole.

After applying "ClearMixer" I think that you can understand by listening to audio (excessive application), but if interference is excessively reduced, it will make a slightly unpleasant sound. I think that it is good to control the amount of interference while looking at the "Total Interference" index of "ClearMixer". For reference, "Total Interference" of the original audio is about 15%, "Total Interference" of the audio after applying "ClearMixer" is about 8%, "Total Interference" of the audio after applying "ClearMixer" is about 3% is.

I think that it is merit that "ClearMixer" can be used to objectively mix while watching indices.

"ClearMixer" latest version download

Demo Version

Product version

Buy product version

* Update method: Run install.bat and it is OK.

* Please see the bundled README for usage.

Summary

We introduced videos and demonstration audio that show the effect of "ClearMixer".

AI Mastering has been updated. Below is the update content.

Mastering delete function

The function to delete mastering was added. You can delete from "Mastering list screen" or "Edit button" on each mastering detail screen. Please note that it can not be Undo. ※ It can not be used in the smart version

Mastering protection function

We added a function to protect mastering and prevent it from being automatically deleted. You can protect up to 100 at the same time with the premium plan limited function. The retention period is the contract term of the premium plan. ※ It can not be used in the smart version

How to use

You can "protect" from the "Edit button" on the mastering list screen or each mastering detail screen.

Whether or not it is protected is indicated by "Status column" on the mastering list screenIn can be confirmed.

Notes

It is not intended to store important data using the protection function. I recommend downloading important data and keeping backups.

Video encoding with fair priorities

We decided to do video encoding with fair priority. Even if there are more masters, it will be hard to be affected.

The mastering process itself had been done with fair priorities, but its priorities were not applied to video encoding. Therefore, when there were someone who mastered, the sound source was completed as soon as possible, but there was a phenomenon that the moving image was slow to be completed. It is a fix to solve it.

Changing display contents of settlement list

The display content of the settlement list on the "User setting" page was corrected.

Specifically, when opening the PayPal link but not completing the settlement, "unpaid" was displayed, but it was hard to understand, so it was not displayed at all in the first place.

Fixed bug where login failed

Fixed a bug that login failed on the desktop version or some browsers. We apologize for any inconvenience. Please download the latest version from the link below.

Download the latest desktop version

We introduce AI Mastering's "limiter error".

What is "limiter error"?

"Limiter error" is the error of sound caused by AI Mastering's limiting process .

Generally, if the sound pressure is raised too much with the limiter, the sound will be distorted, but the limiter error quantitatively expresses that distortion. It is designed based on psychoacoustics so that it approaches the human sense as possible. The unit is dB. If the limiter error is 1 dB or less, you should not be able to hear the difference.

The figure is simplified, but in fact it calculates it in a slightly more complicated way.

Cause of occurrence of "limiter error"

The cause of the limiter error is that the waveform is distorted as a result of attempting to bring the sound pressure closer to the target sound pressure while limiting the peak of the waveform to less than Ceiling by limiting processing.

There are some sounds that are distorting and distortion-proof with limiting. Whether distortion is easy depends on the limiter, but in the case of AI Mastering, percussion instruments are hard to distort, bass, close to sine waves, sustained sounds are easy to distort .

Tips for not generating "limiter error"

Decrease target sound pressure

Lowering the target sound pressure makes it easier to reduce the limiter error. This is the best when you can lower the sound pressure.

"Ceiling mode" is set to "peak"

In AI Mastering's custom mastering, you can choose whether to limit based on True Peak or normal peak using the Ceiling Mode option.

Because it suppresses the peak more conservatively based on True Peak, Ceiling goes down, and limiter error tends to occur.

By setting "Ceiling mode" to "Peak" , limiter error can be reduced. Instead, it is more likely to be distorted by lossy compression.

Small base peak

The bass is a low-pitched sustained sound close to the sine wave, so it's easy to distort at the limiter.

Although it is not limited to AI Mastering, if you use an amp simulator etc. to increase the overtones of the bass , you can reduce the peak while maintaining the audible volume. Instead, the sound quality changes.

I do not care that the "limiter error" is large

There are musically acceptable and unacceptable cases in which "limiter error" occurs.

For example, the sound of the bass is temporarily too large and the peak pops out, so if you lower the sound of the bass overall, "limiter error" will occur but I think that it is musically acceptable.

On the other hand, I think that it is not musically acceptable in the case of distorting it by forcibly suppressing the base peak in the same situation.

"If there is no limiter error occurred, there is no problem" is established (at least though it is designed to be established), "There is a problem if a limiter error has occurred" is not established .

So, it is also a hand that you do not mind that "limiter error" is big if you listen with your ears and have no problem.

Improvement plan "Limiter error spectrogram"

"A limiter error" displayed by AI Mastering is the average of the entire sound source. In reality, the local "limiter error" differs for each time and band.
If there is something like a spectrogram of "limiter error", you can see at which time and within what band "limiter error" is occurring, you can easily identify the cause.

It is under consideration.

Summary

We introduced about AI Mastering's "limiter error".

We introduce PhonicMind, an online service that extracts / removes only vocals from the sound source, and other similar software / services.

What is PhonicMind?

PhonicMind is an online service that can automatically extract and remove vocals from 2-mix sound sources.

When uploading the sound source, you can download two sound sources, the sound source which extracted only the vocal from the sound source and the sound source which removed only the vocal.

PhonicMind official website

Phonic Mind's reputation

Looking at the response of the link below, PhonicMind's reputation seems to be good.

PhonicMind, a vocal remover that actually works when it comes to isolating the vocals. HIGHLY RECOMMENDED! from makingvaporwave

Has anyone tried PhonicMind? from IsolatedVocals

How PhonicMind works

PhonicMind seems to be using deep neural net.

PhonicMind’s vocal remover uses deep neural networks to do vocal elimination.

Source: https://phonicmind.com/faq/

Other vocal removal / extraction software

Vocal Remover (vocalremover.org)

Like PhonicMind, VocalRemover is an online service that can automatically remove / extract vocals from 2-mix sound sources.

You can remove vocals from "Vocal Remover" in the left menu and vocal extraction from "Vocal Extractor".

Vocal Remover

When I tried it, the band is near the vocal and the sound with the localization in the center is extracted together with the vocal. Also, vocal removal has reverb components left.

As you can imagine, I think that you are using the traditional method of extracting vocals using information of localization, frequency band, transient.

VocalRemover (vocalremover.com)

Like PhonicMind, VocalRemover is an online service that can automatically remove / extract vocals from 2-mix sound sources.

VocalRemover

I tried it, the quality is higher than vocalremover.org. The reverb component after vocal removal is also weak. I felt the same quality as PhonicMind.

Lakeside Audio Isola Pro FX

Lakeside Audio Isola Pro FX is a VST plug-in that can extract various instruments semi-automatically from 2-mix sound sources.

Lakeside Audio Isola Pro FX

There are "MIDI mode" giving hints at Midi and automatic mode to specify the frequency band, and "MIDI mode" seems to have higher quality. Because it is VST, it can process in real time.

It is a comparison video of PhonicMind and Lakeside Audio Isola Prox FX. A little artifact has appeared, but I felt the same quality as PhonicMind.

iZotope RX 7

iZotope RX 7 is a standalone software for repairing and adjusting 2 - mix sound sources, which supports both music production and post production.

iZotope RX 7

The initial version of iZotope RX was announced in 2007, with the release of RX 7 in 2018, the function of automatically extracting vocals, bass, percussion etc from 2 - mix sound source and readjusting the volume was added.

According to the following information, it seems that neural net is used for sound source separation algorithm.

The evolution of our intelligent audio technology continues with the Music Rebalance module in RX 7. Music Rebalance is a new tool that gives users the ability to boost, attenuate, or even isolate musical elements from audio recordings. It is a natural progression of our neural network-based source separation technology, first introduced in the forms of Dialogue Isolate and De-rustle in RX 6 and now evolved to extract multiple musical components from complex mixes.

Source: https://www.izotope.com/en/blog/music-production/exploring-the-technology-that-makes-rx-7-music-rebalance-possible.html

Audionamix XTRAX STEMS

Audionamix XTRAX STEMS is a standalone software that fully separates 2-mix sound sources into three vocals, drums, and other instruments.

Audionamix XTRAX STEMS

According to this information, it uses a neural network and seems to be superior to ADX TRAX.

Audionamix ADX TRAX

Audionamix ADX TRAX is a standalone software that extracts vocals. We are specialized in vocal extraction. Unlike PhonicMind, you can fine-tune manually while watching the spectrum.

Audionamix ADX TRAX

BlueLab REBALANCE

BlueLab REBALANCE is a VST that allows you to adjust the volume of each 2 - mix sound source divided into 4 vocals, bass, drums, and other instruments. It is released in January 2019. It is VST so it can be processed in real time.

I tried it, it was lower quality than PhonicMind. I guess, I guess you are using a traditional algorithm.

BlueLab REBALANCE

Which vocal removal / extraction service should be used?

I want to make a karaoke sound source

I think PhonicMind or VocalRemover is good.

Because it is a web service, it is unnecessary to install software and other troubles.

I want to copy (transcription) my ear

I think that Lakeside Audio Isola Pro FX, iZotope RX 7 or Audionamix XTRAX STEMS is good.

Musical instruments other than vocals can also be extracted. I do not know which of these is better because I do not use it.

Summary

We introduced the software / service for extracting and removing vocals on behalf of PhonicMind.