I'm testing and comparing directly against other large industry software that does the same job. And I have to say, I’m impressed!!! This is a beast!!! It does a job on the same level or even better than the more expensive software I have. It doesn’t make the computer unusable like the others and makes it faster. And what impressed me most is that it costs only a fraction of the value of others!
I came across a mention of this program on Reddit today. Vocal isolation works great. As an alternative, someone suggested a "free" 5 GB solution. What the heck? Why should I download a bunch of garbage if everything fits perfectly into 50 Megabytes?
Vocal isolation works well, with no "tails" left over from the music. At the same time, I do not notice any significant frequency limitation in the sound editor.
It's a really good deal to pay once instead of subscribing to Mvsep. Moreover, the quality of vocal isolation is higher than that of the MDX23C.
p.s. thanks for the quick fix to the UI that was stopping me from purchasing.
It worked incredibly well with the different genres of music I have in my library. Can output only clean vocals or only music.
Every other audio editing software I've tried to do the same thing was never that good, you could always hear annoying bits of music along with the vocals. This does an ideal job when creating acapella.
I understand that you are aiming for a compact size, but support for GPU acceleration is still necessary. I was quite pleased with the separation results, but processing on the CPU quickly drains the battery.
At first I thought it was a web installer and it would download damn Python with dozens of libraries, like similar applications do. But a miracle happened - the installer did not go online, but immediately launched the GUI. After the first start, the window looked translucent, as if the image was damaged. After the restart, the interface returned to normal and caused no further problems. The speed is acceptable; splitting a 5-minute track takes less than 2 minutes. The quality of the separation is quite decent, but without saving and studying in an audio editor it is difficult to draw definitive conclusions. I would like to see support for GPU acceleration in the future. Overall, a solid 4.
The UVR installer takes up 1.5 gigabytes, the trained Demucs Vocal model alone takes up 160 megabytes and is inferior in quality. VocalRip takes up less than 30 megabytes, is faster and isolates vocals better. This is impressive!
I agreed to become a beta tester because I hate neural networks made in Python. Judge for yourself, installing a seemingly simple application entails gigabytes of dependencies. The Python310 folder takes up about 10GB of disk space. Add here huge models with hundreds of megabytes and you will understand what I mean. The guys from Abyssmedia promised to keep it to 100MB and even exceeded their promises. I don’t know how the quality of sound separation is assessed, but here it is better than in the popular Demucs. Since I have an AMD video card installed, I cannot use GPU acceleration and for me the operating speed is quite comfortable on the Ryzen 5500. The ability to adjust the volume of the music and the artist separately is very useful for karaoke and allows you to "peep" if you are not getting into the rhythm.
I was also a beta tester, so I rate not only the developers, but also myself, for catching a critical bug :-) When I received the debug versions, their size was more than 100 MB, so when I saw the final release was less than 30 MB, I thought that they forgot to put something there.
I have already tried models from Demucs, UVR, Roformer, BadnSplit, but here the quality of voice separation is noticeably higher. Well, in terms of quality/size ratio, VocalRip is generally unrivaled.