A channel update! (For those wondering what's up and where I'm going) 😎 The development of AI technology over the last few years has been pretty meteoric. While some parts of that frighten me (machine overlords), it has been pretty awesome for the Karaoke community. 'De-mixing' AI has allowed the many dozens of new Karaoke brands to make all these videos in recent years. Basically, using 'AI' to remove the vocals from any song, or isolate backing vocals etc.
For my own part, I've remained extremely conservative about this technology. I began this channel before de-mixing AI was a thing, focusing on stems and instrumental mixes, and have moved very slowly on adopting the new technology compared to my peers. This is because I thought that the quality of the de-mixing AI results just wasn't good enough on most things for a long time. It has gotten substantially better over the last five years though.
I began really adopting it about three years ago in earnest to isolate backing vocals. I mention this because the 'Updated videos' are for sure still a big part of my immediate channel plans. Back when I started all this in 2017, there was no AI to effectively remove vocals or isolate backings vocals. I began the channel working with stems and instrumental mixes, and when there was no backing vocals stem, I just settled for posting the Karaoke without backing vocals. Now that there are AI models to isolate backing vocals, I want to make sure I add them to as many old videos as possible, and include them in all new videos. I've been taking careful notes on this .. and there's a couple hundred tracks that I still can't isolate .. (thinking of some Blink 182 tracks in particular among others) .. so those videos will remain with the old thumbnails and low resolution until they can be fixed. For the rest, I am working on updating them. 'Red Hot Chili Peppers - Otherside' is the most significant recent example of this. There are currently several hundred videos on the channel for which de-mixing AI was used to help with the backing vocals, either from splitting vocal stems or pulling them from the official finished mix and them splitting them.
Second, all of the videos with the little 'Neural' UVR logo in the bottom left hnd corner of the thumbnails were made exclusively with de-mixing models. No original instrumental mixes or stems aiding. There's a 'UVR' playlist on the main page of my channel that lists all of the videos done this way. I've dabbled in this full on approach over the years but never committed to it because I felt there were still too many quality issues with the final results.
Well, the latest and greatest AI model has me sold. It's not perfect .. and there will be better models in the future, but this new model blows all past models out of the water. It's a revelation, frankly. A massive leap forward .. and it produces results that make me want to really dive into using this technology more significantly going forward. This was always going to be the long term future of the channel, and I'm pumped to feel like that future is finally arriving.
First things first though .. I'm going to focus on updating the existing tracks from the UVR playlist for a little while. They are already on the channel and being used and I want to clean up the audio on them. From there, I'll expand out to do tons more stuff I've been wanting to do for a long time. Rolling Stones, AC/DC, Black Sabbath and more.
Also, I've still got thousands of instrumentals and stems collected in many folders and drives that deserve attention. Up front, I've got about 300 more tracks from stems that are mixed and just need the video work finished, so those will get a higher level of priority in the coming months. Everything from Queen to Smash Mouth, Linkin Park, Megadeth and more.
Once I've got the main house cleaning a bit more settled, I plan to FINALLY start taking requests ... I know .. been a long time coming .. but it's in the plans. I need to sort out how and where I'll organize that, but it's coming. I'll update more on that here in the new year :)
CC Karaoke
A channel update! (For those wondering what's up and where I'm going) 😎
The development of AI technology over the last few years has been pretty meteoric. While some parts of that frighten me (machine overlords), it has been pretty awesome for the Karaoke community. 'De-mixing' AI has allowed the many dozens of new Karaoke brands to make all these videos in recent years. Basically, using 'AI' to remove the vocals from any song, or isolate backing vocals etc.
For my own part, I've remained extremely conservative about this technology. I began this channel before de-mixing AI was a thing, focusing on stems and instrumental mixes, and have moved very slowly on adopting the new technology compared to my peers. This is because I thought that the quality of the de-mixing AI results just wasn't good enough on most things for a long time. It has gotten substantially better over the last five years though.
I began really adopting it about three years ago in earnest to isolate backing vocals. I mention this because the 'Updated videos' are for sure still a big part of my immediate channel plans. Back when I started all this in 2017, there was no AI to effectively remove vocals or isolate backings vocals. I began the channel working with stems and instrumental mixes, and when there was no backing vocals stem, I just settled for posting the Karaoke without backing vocals. Now that there are AI models to isolate backing vocals, I want to make sure I add them to as many old videos as possible, and include them in all new videos. I've been taking careful notes on this .. and there's a couple hundred tracks that I still can't isolate .. (thinking of some Blink 182 tracks in particular among others) .. so those videos will remain with the old thumbnails and low resolution until they can be fixed. For the rest, I am working on updating them. 'Red Hot Chili Peppers - Otherside' is the most significant recent example of this. There are currently several hundred videos on the channel for which de-mixing AI was used to help with the backing vocals, either from splitting vocal stems or pulling them from the official finished mix and them splitting them.
Second, all of the videos with the little 'Neural' UVR logo in the bottom left hnd corner of the thumbnails were made exclusively with de-mixing models. No original instrumental mixes or stems aiding. There's a 'UVR' playlist on the main page of my channel that lists all of the videos done this way. I've dabbled in this full on approach over the years but never committed to it because I felt there were still too many quality issues with the final results.
Well, the latest and greatest AI model has me sold. It's not perfect .. and there will be better models in the future, but this new model blows all past models out of the water. It's a revelation, frankly. A massive leap forward .. and it produces results that make me want to really dive into using this technology more significantly going forward. This was always going to be the long term future of the channel, and I'm pumped to feel like that future is finally arriving.
First things first though .. I'm going to focus on updating the existing tracks from the UVR playlist for a little while. They are already on the channel and being used and I want to clean up the audio on them. From there, I'll expand out to do tons more stuff I've been wanting to do for a long time. Rolling Stones, AC/DC, Black Sabbath and more.
Also, I've still got thousands of instrumentals and stems collected in many folders and drives that deserve attention. Up front, I've got about 300 more tracks from stems that are mixed and just need the video work finished, so those will get a higher level of priority in the coming months. Everything from Queen to Smash Mouth, Linkin Park, Megadeth and more.
Once I've got the main house cleaning a bit more settled, I plan to FINALLY start taking requests ... I know .. been a long time coming .. but it's in the plans. I need to sort out how and where I'll organize that, but it's coming. I'll update more on that here in the new year :)
8 months ago | [YT] | 348