Subscribe
Sign in
LALAL.AI 2.0 — World’s first 7-stem separation service for audio & video
LALAL.AI is an online service for music source separation powered by AI technologies.
Extract vocal, instrumental, and various musical instruments from any song or even video.
31
Replies
Marcel Weiss
Nice!
Share
✅ Eager to try it. If this thing actually works, I'd be amazed! … would be so so sick for remix.
@bramk Thanks for hunting us ;-)
Hey Hunters!
Just a year ago, LALAL.AI debuted on ProductHunt as an advanced AI-powered service for splitting audio into vocal and instrumental tracks.
Today, we’re excited to present the service as the world’s first 7-stem separator!
You can now use LALAL.AI to extract or remove the following:
🎤 Vocal/voice
🎶 Instrumental
🪕 Acoustic guitar
🎸 Electric guitar
🥁 Drums
🎼 Bass
🎹 Piano
📈 IN-HOUSE DEVELOPMENT
The state-of-art technology we use for stem splitting is 100% original. We developed it from scratch and trained the neural networks ourselves, operating completely independently.
⚙️HOW IT WORKS
The inner workings of LALAL.AI are complex, but the service is super easy to use:
- Select the stem isolation type.
- Upload your audio or video file.
- Listen to the stem previews.
- Sign up & download the full stems.
🤑 SPECIAL OFFER
A free trial is always available for all users.
However, we wanted to treat Hunters to something special for the occasion!
Get a limited time discount coupon for 50% OFF all paid packages.
Claim it before it’s too late!
P.S. You can grow your business with LALAL.AI! We offer B2B solutions that can be installed on your server/AWS or operate from our end. One-time deals and committed cooperation options are available.
This is easily the best service available for splitting anything. What's unique is they allow acoustic / electric guitars to be split as well. The results it gives are so clean! As with any of these services, you can't expect perfection, but I have to say, Lalal is the closest you're gonna get. With time, it's only going to get better, and sometimes the stems it outputs already could be mistaken for the real multi-tracks (if you use them in a remix or something along those lines).
Seeing a service which resamples and enhances low-quality music (22050hz and below, with lossy artifacts) would be nice as well. Just an idea! ;)
@jh9918 Hi Jacob, thank you for your review, we are going to train the network for instruments further and we hope it will give us better results, stay tuned! Concerning your idea, yeah we are thinking about the same to enhance the old stereo recordings or low quality music because this is the most logical and close application of our technologyю
It's astonishing how well this works. It doesn't seem to harm the rest of the music once the vocals are extracted. It's great for singer doing vocal practice, because you can import the vocal-less tracks into a DAW and do your own version of the vocal. And because it extracts isolated vocals, you can really listen carefully to vocal performances that you love to borrow techniques.
@bruce_johnson2 Thank you for your kind words, Bruce! Yes a lot of people use our service for music and vocal learning.
Worked perfectly for the first few vocal removals, but then it started to keep some vocals
@ismail_ghedamsi thanks for feedback, you can always try to upload the better quality of initial file, I mean in flac or wav format, the result can be better then.
@jean_sean Yeah, it's always a compromise between the quality and the processing power, we need to run our neural network on powerful amazon servers otherwise it will be a big loss in quality of splitting.
@jean_sean @alexandramalis Sounds like a collab with Shazam would allow you to cut costs as I believe matching to a song is way cheaper than running ML to split audio. Hope you'll find a way! 🤞
@anatolinicolae Interesting idea, Anatoli, I will share it with our dev team! That's true that the song identifying could help a lot!
The best tool for separating music.
I've tried many similar tools for vocal extraction, but this one left positive emotions, Lalai.ai works amazingly compared to other competitors, plus it gives free access, I always recommend this site to my friends. In addition, it is able to separate more better tools, which other developers do not give. I am satisfied with this service, the best of the best. To the developers of this wonderful service, I wish inspiration and new heights!
@og1 Thank you so much for such positive feedback! It means we are on the right way ;-)
I'm a robotics engineer and also make music as a hobby. I've explored Spleeter, but this is definitely much better and cleaner. Keep up the great work. I'm sure this will just keep getting better the more diverse data you gather and train on.
@geeve_george Yeah, we outperform Spleeter by Deezer and others in terms of quality, and we will continue to improve it!
This is awesome. Lalal.ai does a better job than the other isolation websites out there. Those output isolation that cuts a lot of frequencies, had metallic artifacts, and low quality; while Lalai.ai outputs more natural-sounding cleaner isolation, and it works with almost any song. Being trained on a 20TB dataset is impressive, and good thing they expanded the options to more separate instruments, tho the other isolation options than the Vocals/Instrumental choice, needs improvements cause it isn't clean enough yet. Totally would recommend!
Of all the tools to split vocals from the rest of the track, this is the best -- magical! For some it can provide a tool to escape poverty, by remixing their generation's anthems and even renewing previous ones.
This will provide happiness and joy again in music, where the old tunes had worn out.
Seems to dynamically reconstruct the original, almost like creating a new piece -- especially if working in a lossyless format (I prefer .wav).
Tip: To keep the vocals in to an extent, select the "Mild" processing setting.
@galacticdj Hey Corey, yeah we are thinking to launch this kind of product for the old stereo recordings, stay tuned!
I personally recommend this state of the art separation powered by AI technologies!!