Podcast: Play in new window | Download (Duration: 13:49 — 13.0MB) | Embed
Subscribe: Spotify | TuneIn | RSS
Today we look at the new iZotope RX 8 as a tool for extracting stems from tracks for remixes and mashups. We’ll compare it with Spleeter, some free software made available by Deezer, using a couple of my songs so you can hear how the two compare. I’ve also got an update on Facebook who clarified their recent music guideline changes and what they meant by “music experiences” for live streaming.
You’ll Learn
- My thoughts on RX 8 for separating a track into stems
- Alternatives for creating DIY acapellas and stems
- What Facebook means by “music experiences”
Resources
- PLP 005: Finding Acapellas and Creating Cover Art
- iZotope RX 8
- Releasing Spleeter: Deezer Research source separation engine
- Splitter.ai – AI powered audio extraction – Free! Worked great. Careful with Terms of Service.
- Moises.ai – another web-based Spleeter implementation, but requires an account and offers premium features. Not tested.
- XTRAX STEMS – Not tested, subscription model.
- Spleeter M4L patch – not tested, requires some effort to set up.
- The truth about LIVE-STREAMING on Facebook (Musicians, DON’T panic!!!)
- Facebook blog post on Guidelines for Music (Updated September 11th)
- House Ninja Mashups on YouTube
Transcript
What’s up Heroes, welcome to episode 38 of the Producer Life Podcast. Today we’re going to look at the recently released iZotope RX 8 as a tool for extracting stems from tracks for remixes and mashups. We’ll compare it with Spleeter, some free software made available by Deezer using a couple of my tracks so you can hear how the software performs. Additionally, I’ve got some updates on Facebook who clarified their recent music guideline changes and what they meant by “music experiences” for live streaming.
But first, cue the intro music.
I’ve always loved a good mashup and I talked way back in episode five about places to get acapellas. But what if you can’t get a pristine studio acapella as part of an official remix competition, or find a DIY acapella on one of the various acapella sites? How do you create your own?
iZotope just launched RX 8, which is really focused on audio repair. It has a host of powerful tools to de-ess, de-hum, remove plosives, fix guitar squeaks, remove mouth clicks, and generally make audio files sound mo-better. There are three different versions, elements, standard, and advanced. I’m going to focus on one aspect which is available in the standard and advanced versions, the “Music Rebalance” which is designed to take a master recording and re-balance it by adding more emphasis to a particular part of the mix…. But it can be equally powerful for separating out the vocals, percussion, base, and other elements of the mix for a remix or mashup.
I’m going to use two of my recent tracks, This Way and Ven Pa’ Ca, which are vocal-heavy tracks to give you some actual examples of how this software performed. We’re going to compare that with Spleeter, which is open source software developed by Deezer which is free to use. Unfortunately, while Spleeter is free to use it doesn’t come with a clean, easy to use software front end… so I used a website called Splitter.ai which created a super-simple and easy to use web-based front end. Splitter.ai allows you to upload your tracks and then quickly delivers five stems, all free and without an account requirement.
So let’s start with RX 8. RX 8 has two different ways you interact with it. It comes with VST plugins which provide much of the functionality directly in your DAW. I’m using several of those on this podcast to improve the audio quality, including the De-ess, De-Mouth Click, and Spectral-De Noise.
There is also a stand-alone program which is what I used to separate these files. I dragged my MP3s in for both of these tracks, told it to separate the tracks on the gentle setting, and then let it do it’s thing. Each track took about about 20 minutes. Of note, the Splitter website was shockingly fast, spitting out stems in less than 5 minutes. Splitter offers two-stem and five-stem options, but unlike RX8 it don’t give you the ability to control how strong the separation will be.
Let’s start with my track This Way. Here’s the original final chorus:
Here are the vocals separated by RX8 under gentle default settings
Here are the vocals via Splitter.ai
I felt like these two were pretty close. Each had some artifacts, although different ones, but both did a good job isolating the vocals from the rest of the mix. Absent a studio acapella, I feel like I could work with this. The advantage with RX8 is that I could go back in and tinker with some of the settings. Splitter.ai doesn’t give any options to change settings…. You get what you get.
Let’s listen to the percussion. Here it is again from RX8, again under gentle separation settings.
Here’s the percussion from Splitter.ai.
RX 8 had a clear advantage on this one with Splitter doing something weird with filtering the hats.
Let’s listen to one more of the stems, this is categorized as “other” in This Way.
This wasn’t so clean in RX 8, but then again it’s RX 8’s catch-all category of “other” so that’s sort of expected.
Splitter has this categorized as “Piano” and as you can hear it’s a lot more focused.
Lets try one more example. Here’s the chorus from my track Ven Pa’ Ca
Here are the vocals separated by RX8 under gentle settings
Here are the vocals via Splitter.ai
I preferred the RX 8 version which sounded like it had a more stable gain level.
Let’s listen to the percussion. Here it is again from RX8, under gentle settings.
Here’s the percussion from Splitter.ai.
Both of these were pretty good. I think some of the artifacts could be cleaned up pretty well with a noise gate.
Finally, let’s listen to the “other” category, first from RX8: RX8 got the horns really well, although you’ve got noticeable bleed from the vocals which are in the same frequency.
For Splitter, you’ll notice it completely skips the guitar in the first three seconds and then it does successfully manage to bring out just the piano although it’s very modulated.
Overall, I was really surprised how well Splitter.ai worked. Did I mention it’s free? If you only need an occasional acapella you can’t find or buy online, this could very well be your go-to. Note, however, that usage policy specifies you’re not supposed to use this for tracks you don’t own, so if you’re using it for a bootleg remix you’re taking your chances.
There are other options out there. XTRAX Stems is another option. This software separates tracks into four stems and reviews seem pretty good online. I didn’t test this software because I don’t care for their subscription model of $60/year and didn’t want to have to cancel after their free trial. I’m also a big fan of iZotope software and was already interested in RX8 for this podcast.
RX8 is the most expensive option though. Full price for Standard, the version you need if you want access to Music Rebalance, is $399 although it’s currently on sale for $299. However, if you’re a current iZotope customer if you login to your account you should have a cross-grade for RX8 Standard for $199. Also… I discovered if you show “exit intent” by moving your mouse like you’re going to close the shopping cart during checkout, they’ll offer you another 5% off or about $10. So… $189 with the cross-grade and mouse fake-out. Keep in mind, while this is really expensive if you’re ONLY using it to isolate acapellas… RX8 is full of useful tools for producers and recording artists. I suspect you’ll find yourself using a lot more of them, especially if you’re working with other musicians and doing a lot of collabs. iZotope does offer a 30 day trial on the software if you want to try it. You can’t export your stems, but the VST plugins work fine and you can save projects in your DAW.
Last week I talked about Facebook’s big Terms of Service change effective 1 October, and I also stopped livestreaming my weekly DJ sets to Facebook. Facebook created a bit of panic in the music community, and they’ve since clarified what they meant.
Just to recap, Facebook seemingly updated it’s terms of service, effective 1 October, which gave the appearance they were going to crack down hard on livestreaming musicians. You can go back and listen to episode 37 if you’d like a little more detail. The key phrase in their music supplement which got so many people spinning was, “You may not use videos on our Products to create a music listening experience,” but Facebook really didn’t define what a “music listening experience” was.
CD baby did a really good livestream yesterday after talking with their Facebook contacts, and they provided some tips and advice for musicians. I’ll link their article in the show notes at Producer Life Podcast, but the bottom line is that Facebook is not Spotify. You can’t just stream music with a static cover onto their platform, especially if it’s not your music. Facebook wants engagement, not passive listening.
Here’s a quote from Facebook’s recently updated blog post on music in video:
Our partnerships with rights holders have brought people together around music on our platforms. As part of our licensing agreements, there are limitations around the amount of recorded music that can be included in Live broadcasts or videos. While the specifics of our licensing agreements are confidential, today we’re sharing some general guidelines to help you plan your videos better:
Music in stories and traditional live music performances (e.g., filming an artist or band performing live) are permitted.
The greater the number of full-length recorded tracks in a video, the more likely it may be limited (more below on what we mean by “limited”).
Shorter clips of music are recommended.
There should always be a visual component to your video; recorded audio should not be the primary purpose of the video. Source
So what can we take from this as DJs and performers? Facebook is using algorithms to screen not just the audio, but apparently the video content too. If you’re not moving around your stage, interacting with your audience, and playing your own original content you raise your chances of having your video muted or worse. If you’re a DJ playing mostly other people’s music, your safest options from a copyright perspective remain Mixcloud Live or YouTube.
Thanks so much for listening, if you haven’t already please make sure you subscribe to the podcast and leave a rating and review.
Until next week, this is the House Ninja reminding you to be someone’s Hero today.