Skip to content
Waxy.org
About
Mastodon
Contact

Fast and Free Music Separation with Deezer’s Machine Learning Library

Posted November 4, 2019April 14, 2020 by Andy Baio

Cleanly isolating vocals from drums, bass, piano, and other musical accompaniment is the dream of every mashup artist, karaoke fan, and producer. Commercial solutions exist, but can be expensive and unreliable. Techniques like phase cancellation have very mixed results.

The engineering team behind streaming music service Deezer just open-sourced Spleeter, their audio separation library built on Python and TensorFlow that uses machine learning to quickly and freely separate music into stems. (Read more in today’s announcement.)

The team at @Deezer just released #Spleeter, a Python music source separation library with state-of-the-art pre-trained models! 🎶✨

Straight from command line, you can extract voice, piano, drums… from any music track! Uses @TensorFlow and #Keras.https://t.co/e4lyVtT2lR pic.twitter.com/tDsBMSYiJD

— 👩‍💻 Paige Bailey @ 127.0.0.1 🏡 #BLM (@DynamicWebPaige) November 2, 2019

You can train it yourself if you have the resources, but the three models they released already far surpass any available free tool that I know of, and rival commercial plugins and services. The library ships with three pre-trained models:

  • Two stems – Vocals and Other Accompaniment
  • Four stems – Vocals, Drums, Bass, Other
  • Five stems – Vocals, Drums, Bass, Piano, Other

It took a couple minutes to install the library, which includes installing Conda, and processing audio was much faster than expected.

On my five-year-old MacBook Pro using the CPU only, Spleeter processed audio at a rate of about 5.5x faster than real-time for the simplest two-stem separation, or about one minute of processing time for every 5.5 minutes of audio. Five-stem separation took around three minutes for 5.5 minutes of audio.

When running on a GPU, the Deezer team report speeds 100x faster than real-time for four stems, converting 3.5 hours of music in less than 90 seconds on a single GeForce GTX 1080.

Sample Results

But how are the results? I tried a handful of tracks across multiple genres, and all performed incredibly well. Vocals sometimes get a robotic autotuned feel, but the amount of bleed is shockingly low relative to other solutions.

I ran several songs through the two-stem filter, which is the fastest and most useful. The 30-second samples are the separations from the simplest two-stem model, with links to the original studio tracks where available.

🎶 Lizzo – “Truth Hurts”

Lizzo (Vocals Only)
Lizzo (Music Only)

Compare the above to the isolated vocals generated by PhonicMind, a commercial service that uses machine learning to separate audio, starting at $3.99 per song. The piano is audible throughout PhonicMind’s track.

🎶 Led Zeppelin – “Whole Lotta Love”

Led Zeppelin (Vocals Only)
Led Zeppelin (Music Only)

The original isolated vocals from the master tapes for comparison. Spleeter gets a bit confused with the background vocals, with the secondary slide guitar bleeding into the vocal track.

🎶 Lil Nas X w/Billy Ray Cyrus – “Old Town Road (Remix)”

Lil Nas X (Vocals Only)
Lil Nas X (Music Only)

Part of the beat makes it into Lil Nas X’s vocal track. No studio stems are available, but a fan used the Diplo remix to create this vocals-only track for comparison.

🎶 Marvin Gaye – “I Heard It Through the Grapevine”

Marvin Gaye (Vocals Only)
Marvin Gaye (Music Only)

Some of the background vocals get included in both tracks here, which is probably great for karaoke, but may not be ideal for remixing. Compare this to 1:10 in the studio vocals.

🎶 Billie Eilish – “Bad Guy”

Billie Eilish (Vocals Only)
Billie Eilish (Music Only)

I thought this one would be a disaster—the vocals are heavily processed and lower in the mix with a dynamic bass dominating the song—but it worked surprisingly well, though some of the snaps bleed through.

🎶 Van Halen – “Runnin’ With The Devil”

Van Halen – “Runnin’ With The Devil” (Vocals Only)
Van Halen – “Runnin’ With The Devil” (Music Only)

Spleeter had a difficult time with this one, but still not bad. You can compare the results generated by Spleeter to the famously viral isolated vocals by David Lee Roth, dry with no vocal effects applied.

Open-Unmix

The release of Spleeter comes shortly after the release of Open-Unmix, another open-source separation library for Python that similarly uses deep neural networks with TensorFlow for source separation.

In my testing, Open-Unmix separated audio at about 35% of the speed of Spleeter, didn’t support MP3 files, and generated noticeably worse results. Compare the output from Open-Unmix below for Lizzo’s isolated vocals, with drums clearly audible once they kick in at the 0:18 mark.

The quality issues can likely be attributed to the model released with Open-Unmix, which was trained on a relatively small set of 150 songs available in the MUSDB18 dataset. The team behind Open Unmix is also working on “UMX PRO,” a more extensive model trained on a larger dataset, but it’s not publicly available for testing.

What Now?

Years ago, I made a goofy experiment called Waxymash, taking four random isolated music tracks off YouTube, and colliding them into the world’s worst mashup. But I was mostly limited to a small number of well-known songs that had their stems leak online, or the few that could be separated cleanly with channel manipulation.

With processing speeds at 100 times faster than real-time playback on a single GPU, it’s now possible to turn all recorded music into a mashup or karaoke without access to the source audio. It may not be legal, but it’s definitely possible.

What would you build with it? I’d love to hear your ideas.

Thanks to Paige for the initial tip!

Updates

This thing is dangerously fun.

nobody should have this kind of power pic.twitter.com/4vbl2MGK4Z

— Andy Baio (@waxpancake) November 5, 2019

November 11. You can now play with Spleeter entirely in the browser with Moises.ai, a free service by Geraldo Ramos. After uploading an MP3, it will email you a link to download the stems.

Also, the Deezer team made Spleeter available as a Jupyter notebook within Google Colab. In my testing, larger audio files won’t play directly within Colab, and will need to be downloaded first to listen to.

25 Comments

Panic Announces Playdate

Posted May 22, 2019April 14, 2020 by Andy Baio

Today, Panic announced Playdate, their lovingly retro-modern handheld gaming system with a lineup of custom-made indie games from some of my favorite game designers in the world.

I’m fortunate enough to know the Panic folks, some of the kindest and most creative people I’ve ever met, and I’ve followed this project over the last few years—almost since its inception.

Designed and engineered in partnership with Teenage Engineering, Playdate features a unique crank in addition to standard directional arrow and button controls, which enables new kinds of experimental gameplay.

Introducing Playdate, a new handheld gaming system from Panic.

It fits in your pocket. It's got a black and white screen. It includes a season of brand-new games from amazing creators. Oh and… there's a crank???? https://t.co/WiIPUkpjSq

Yes. A thread… pic.twitter.com/47BwSOtiiP

— Playdate (@playdate) May 22, 2019

Best of all, the game lineup is a complete surprise and comes with the purchase of every Playdate: a dozen secret games released weekly in the first “season,” automatically downloaded to the device every Monday, from a team of all-star indie designers including Keita Takahashi, Bennett Foddy, Zach Gage, and Shaun Inman.

About those games. We reached out to some of our favorite people, like @KeitaTakahash, @bfod, @helvetica, @shauninman, and many more.

Here's a peek at one: Crankin's Time Travel Adventure, from Keita. It's fun and funny. pic.twitter.com/0Ibwqr5k3I

— Playdate (@playdate) May 22, 2019

It’s very special, and I’m so excited to see it finally announced after years of secret development. I can’t wait to buy one.

Dad

Posted February 5, 2019April 14, 2020 by Andy Baio

My father passed away on January 24, less than two weeks ago, while my family was still reeling from the news that he had developed Stage IV lymphoma.

I’ve never talked about my dad here before, mostly because I don’t write often about my personal life, but I’ve made exceptions in the past to commemorate and honor the deaths of people who meant a lot to me.

On Saturday, we gathered together at the Catholic church in Burbank where my parents were married and my grandfather’s funeral was held, and we said goodbye to my dad, Stephen Baio.

At the service, I read the eulogy below, which tried to explain why he was special to so many of us. Writing it was easy, at least in part because he wrote the best bits himself. The hard part was reading it.


There are three words I keep hearing over and over again since my dad passed away, when I’m talking to people who knew him — some I know well, some who haven’t seen me since I was a little kid, and some of you that I’d never met before.

Those three words are, “Everyone loved Steve.”

It’s true, everyone loved Steve. To know my dad was to love him.

Stephen Baio was impossible not to like. He lit up a room with a joyful and contagious enthusiasm — sweet and sincere, and genuinely funny. He was unabashedly silly. He laughed at his own jokes and at yours. He made it very easy to love him.

He was generous to everyone he met, whether it was someone he knew for 30 minutes or 30 years. To me, it always felt like he had a hard time saving money because of how much he loved spending it on others.

My dad made people happy, and as long as you’re making yourself and other people happy and you’re not hurting anybody, who cares what anybody else thinks? You’re probably doing the right thing.

My dad taught me the importance of being yourself. Being true to who you are. He taught me the values of loyalty, friendship, and kindness.

And these weren’t lessons that could be spoken: he taught me by how he lived his life, it was central to how he walked through the world, and how over time, you change it by who you touch along the way, in ways big and small.


After he passed away, I went into his bedroom to find some photos for this memorial service.

And in one big box, I found photos from his entire life. Photos from his childhood with his mother, Teresa, and with his brother, Anthony, and his late sister, Nicolina, my Aunt Nicki.

I found photos with old friends in the ‘70s, camping in the mountains and just goofing around. Photos from his prom, from parties, from his wedding and with my mom. So many from my childhood, going to Disneyland and the beach, dressing up for Halloweens, Christmases, and Easters long ago. Photos from the tours he went on, traveling with bands or shooting photos of monuments in Tokyo. And later, photos from my own wedding, photos with his grandson.

But after that, I found three more big boxes of the same size, and I was surprised to see them stuffed full of every greeting card, postcard, letter, and memento he ever received. Clippings from newspapers, birth announcements and obituaries, matchbook covers and cassette recordings of long-forgotten conversations. Piles of memories spanning decades, immaculately organized into dozens of manila envelopes, filled to bursting.

It was such a clear reminder how much every one of your friendships and relationships meant to him.

Everyone loved Steve, and you should know without any question or doubt, that he loved you all back.


Seven years ago, my Dad wrote a long handwritten letter, one of thousands he’d written in his life. But this one was special, it was only meant to be read in the event of his passing.

I read it for the first time the day after he died, and I’d like to read from it now.

To Whom It May Concern,

I guess this would be considered my last Will and Testament.

Anyways, whatever it is, these are my wishes, for things to be done whenever I’m gone.

I’m hoping there’s not a lot of things to ask for, I don’t want to be too much of a pain in the ass.

So here goes, I’m writing these things down as I’m writing and thinking.

He goes on to talk about how he wanted his personal belongings distributed or donated. He talks about how he wanted to be cremated, and how he wanted his ashes scattered in the High Sierras, up by Olancha, Mammoth, and June Lakes, places that he loved so much.

He said that he didn’t want an expensive funeral, just a memorial with all of his friends and family through the years — all of you — to come together and have what he called “a nice party.”

He wrote, “I want all the people who meant a lot to me to be there. I’m sure you’ll have a good time.”

I’m going to give my dad the last word, by reading from the end of his will, a message that he addressed directly to all of you — the people who meant so much to him in his life. Here’s what he wrote:

Well, I think that’s about it for now. If I think of anything else, I’ll let you know. (Ha Ha.)

I’ll sure miss you a lot, more than you’ll ever know.

Maybe I’ll see you where ever God decides to send me. I hope there’ll be fishing and pretty girls, good music, my Mom’s Italian spaghetti, and beer.

Thank you all for everything, your love, your friendship, good fun, good laughs and memories, and for the pleasure just to know each and every one of you.

With all my love, now and forever.

Bye for now,

Me
Stephen

Thank you.

9 Comments

Suck.com, Gone for Good (For Good)

Posted January 11, 2019April 14, 2020 by Andy Baio

13 years ago, I wrote a post here called “Suck.com, Gone for Good” when the seminal webzine appeared to go offline in December 2005, redirected to a porn search portal.

Fans quickly rallied around it, with some scrambling to put up mirrors and archives, including Suck contributor Greg Knauss and the dearly-missed Aaron Swartz. A few days later, Lycos apparently changed their mind and pointed the domain back to its original server, where it managed to stay for 13 more years.

The site’s had near-death experiences in the past, including some downtime in 2015, but it’s somehow managed to stay online at its original domain for 23 years — more than 17 years after its last column on June 8, 2001.

Three weeks ago, on December 16, 2018, the ownership changed hands from Lycos to a private GoDaddy account. Shortly after that, the Suck.com domain started redirecting to this Weebly account with an “under construction” message.

Yesterday, I noticed the content changed, now displaying a series of sports-related polls, and the following greeting:

Welcome to Suck.com. All sports, all the time. Let your voice be heard! You tell us, does it suck or not?

I reached out to the new owners, but no response yet.

Assuming the domain was sold legitimately by Lycos, this permanently spells the end of Suck.com.

For retrospectives on Suck’s role in early web history, these links do a good job:

  • Josh Quittner’s 1996 Wired article about the site’s origins.
  • Keep Going’s “The Big Fish,” a 2005 retrospective about the impact of “the first great website”
  • Engadget’s Nicole Lee interviewed the Suck crew after their appearance at XOXO 2015.
  • The Internet History Podcast with Suck founders Carl Steadman and Joey Anuff, and the followup episode.
  • In 2016, The Atlantic’s Anna Wiener wrote about “the best magazine on the early web.”
  • In 2017, The History of the Web covered “The Web After Suck” in its second post.

If you’re feeling nostalgic, Suck, Again is an ongoing newsletter by Mark Macdonald that emails each column to you, 20 years to the day.

Sadly, every existing link to Suck’s articles and features are now broken, but of course, available in the Wayback Machine.

If you want to keep it alive for a while longer, add the line below to your hosts file and Suck.com links will still work for as long as Lycos keeps the server online.

209.202.254.90 suck.com

Or use Greg Knauss’s proxy, which is live once again at suck.eod.com.

Thanks to Carl Steadman, Joey Anuff, Heather Havrilesky, Tim Cavanaugh, Ana Marie Cox, Terry Colon, Owen Thomas, and every other Suckster for making something special we’re still talking about all these years later.

Leave a comment

Morph Cut Transitions on TV News

Posted December 13, 2018April 14, 2020 by Andy Baio

Yesterday, this clip from BBC News popped up on Twitter, in which a small child appears to materialize in the background of a woman-on-the-street interview.

WTF… does anyone else see the child teleport? pic.twitter.com/P0ju9J9cby

— @realTewkesburyBC (@TewkesburyLeak) December 12, 2018

If you watch the woman’s face at the same time the boy appears, you can see her expression morph into a smile.

This technique is known as a Morph Cut, a feature added to Adobe Premiere Pro in 2015, intended to smooth transitions in interview footage, removing unwanted pauses, stutters, and filler words (“like,” “um,” and “uh”) without hard splices and cuts.

The results, when used appropriately in interview footage without a changing background, can be nearly seamless.

It’s likely that BBC News used a morph cut in the clip above to tighten up the interview without changing its meaning. But it’s also ripe for abuse and fully capable of altering the meaning of an interview, and in many cases, undetectable.

I’ve known radio interviews were edited like this for years, but the BBC News clip is the first time I’ve seen the technique used in a video interview… or is it?

How many times have you watched footage that was subtly modified using off-the-shelf software, and never knew? Would you ever notice? Would you care?

3 Comments
⇠ Older Posts
Newer Posts ⇢
Waxy.org | About