Vocal elimination is sought after by DJs, karaoke lovers, sound producers, professional musicians and enthusiasts. It’s fairly easy to find separate acapella and instrumental tracks of popular songs, and even some of the more underground artists release isolated stems along with the final mix. However, if you want to play around with the separated vocal and instrumental tracks of everything that’s in-between mainstream and obscure, you can’t go without a specialized tool.
Vocal isolation is a notoriously difficult task that never yields 100% perfect results. Due to the vocal positioning, mastering peculiarities and errors, and the nature of the process itself, the outcome will always have some vocal residue and other artifacts. But with center-panned voice, stereo sound high-quality audio, and a good separation technique, it’s possible to get decent results. Here are 5 online services designed to remove vocal stems from any song.
As stem separation remains a tedious process, artificial intelligence is gaining momentum as a modern, easier solution. Machine learning algorithms used in the emerging AI-powered services are automatically improving through experience. The more songs they are fed, the more refined isolation they provide.
LALAL.AI was trained on 20TB of data with 45 million neural network parameters to achieve superior results in comparison with other members of the AI-wave. The service allows up to 100MB per upload — a whopping number in contrast with only 10 to 30MB usually provided by others.
On top of that, it supports all audio formats, plus outputs the separated stems in the same format as the songs you input. Most importantly, LALAL.AI’s vocal removal precision is markedly better than that of your average source separation software. The results, both acapella and instrumental stems, can always be tweaked to your individual sound perception and taste with the help of the processing filters.
PhonicMind is allegedly the first vocal isolator based on artificial intelligence. Just like LALAL.AI, it’s not only a vocal remover but also a stem separator. PhonicMind splits an uploaded song into four stems — vocals, drums, bass, and one for everything else. Supporting AAC, MP3, WMA, FLAC, WAV, and AIFF, the service allows uploading songs under 9 minutes and 30MB for trial, and spits out a 30 seconds-long preview after a couple of minutes of processing. What’s strange, is that PhonicMind doesn’t offer the first 30 seconds for quality evaluation but rather randomly decides which part of the song is available as a preview.
As for the separation quality, PhonicMind does a better job creating karaoke versions than acapella. Vocals are isolated from other instruments nicely but have sensible echo and distortion effects thrown onto them. Instrumental is good overall but vocals are not completely canceled. In their place, there is some muddy foreign sound resembling distorted barking that appeared in the majority of tested songs.
VOXISO is also a member of the AI source separation wave. The name literally stands for “vocal isolation” (“vox” = vocals, “iso” = “isolation”) but the range of services VOXISO offers is not limited to only this single task. It’s a multifeatured online tool that allows for instrumental separation, acapella creation and isolation of individual instruments. It supports a wider variety of formats, including MP3, WAV, FLAC, WMA, OGG, AAC, AIF, MPA, MPC, OGA, IFF, M4A, and M4B. The file uploading and processing speed is almost lightning fast, never exceeding one minute.
However, the file size is significantly more limited than that of the previously mentioned vocal removers — only 10MB, and that isn’t just trial version restriction. Basic plan subscribers are limited to the same file size uploads while paying $5.99 a month. The vocal and instrumental stems separation is comparable to PhonicMind, with voice being almost entirely canceled, still leaving a hissing noise in its place in the instrumental.
4. Vocal Remover
Another online tool with a self-explanatory name, Vocal Remover offers a refreshingly free model of operation and the widest selection of features among the considered services. On top of isolating and removing vocals, it offers pitch and tempo changing, audio cutting and joining, voice and karaoke recording, as well as audio files conversion, and even BPM (beats per minute) detection. It’s a full-fledged audio editing online service that can be used for many different purposes.
Vocal Remover supports all popular audio formats and doesn’t seem to have any strict file size limitations. The song length is the only restriction — you can’t upload files that are longer than 10 minutes which is adequate given the majority of popular songs average on 3 to 5 minutes. Separated vocals are impressive, could even be flawless if not for the slight staggering after high notes. The faults of karaoke versions produced by Vocal Remover are pretty much the same as shown by the previously mentioned services.
Unlike the preceding vocal removers in this list, XMinusPro has its own vast collection of songs to choose from. Users are also free to upload tracks from their own music collection with no limitations on the length and size — there are no explicit restrictions stated on the site and it ate an hour-long 128MB track like it’s nothing.
Since the vocal removal is not as automated as in AI-powered tools, XMinusPro gives a lot of room for manual adjustment and experimentation. It supports both stereo and mono, allows setting the sample rate (choosing from 32kHz, 44kHz and 48kHz), changing the vocal panning and cutoff frequency, as well as tremble tuning. The service creates a pretty subpar karaoke track on basic settings with noticeably muffled drums, but with a few twists and turns, user can achieve decent results.