I have been part of this community for a long time. To be honest, it's my main source of music information, along with a sister site, for longer than I have been a member. So I chose to share what I built here first.

I did want to share it with the community to which I have been the closest. But I am starting to realize it's not as friendly around here as I thought. I wanted this to be a give-a-penny-take-a-penny type situation. I built a tool I needed and use often in my workflow. I wanted to make it easy for my friends and the communities I'm part of to help and benefit from this. As for going commercial, it didn't cross my mind, but I'm realizing that maybe I could use some funding to deal with all of these issues faster and easier. The only way to do that is to hire more people. So, a commercial solution might come to help me with stress and workload.

This isn't my main job. I am a touring DJ under the name Flosstradamus (http://www.soundcloud.com/flosstradamus). I have multiple gigs weekly. So, when I'm not traveling, my weekdays combine my real life, music, and advising jobs (Suno, Soundful), working on Rubin, Weights, Replay. This isn't the only space where I am squashing bugs. (We have a Discord). I just got home today from my gigs over the weekend. So, I am communicating on here today.

There are a lot of reasons, one being the above. Secondly, my team and I at Replay, Rubin, Weights are constantly fixing all the issues that come across our plate. You can scroll up and see how many we are getting in here. Add all the other forums like Reddit, Discord, etc., where we have to deal with issues. It all adds up, and we are pushing updates multiple times per day. We fix bugs and keep moving. We are adding a change log along with a new look overall in the new year.

This is a passion project we are doing for FREE but you are making me think that We need to charge and hire in some help.

I hope this helps you better understand me. Also, it might help you think about any other human in a position of a FREE product they are building. We all have our own lives. Your life isn't mine and mine isn't yours. Take a sympathetic approach when it comes to other humans who create and build.

As for the main body of your message, I've seen this argument before, but what we're building over at Suno (http://app.suno.ai) has already blown me away. You can see the progression of DALL-E 1 vs. 2 vs. 3. Prompt-to-image has grown so much in such a short period of time.

What we have at Suno with Chirp is the DALL-E 1 of where it will be in 365 days.

I prompted to make a song about AudioSex in the style of Frank Sinatra (https://vocaroo.com/1gj6tUKdIlmt).

Also, Rodger mentioned in this forum and has shown many examples of Suno X Replay in combination. We will be creating great music sooner than later.

Also, @Rodger mentioned in this forum and has shown many examples of Suno X Replay in combination. AI will be creating great music sooner rather than later.

 

Last edited: Dec 13, 2023

What do you mean? This is fully running on a server with high powered GPUs. The compute behind this isn't really available to the consumer. Im confused on what you mean?

It also runs through AI CUDA cores and as of today's date there is no way to run CUDA on mac.

 

Yeah sadly. The tech behind the tech behind Replay (on mac) Needs Apple's new MPS GPU structure to run. It's starting to feel like the more computational plugins, apps, and software are taking advantage of the new structures of compute.

 

Marie and Per Back together one more time Thanks to https://www.tryreplay.io/

 

Hello! Everytime i run the app i get this error. Any hint? it seems a very nice app and I would like to be able to try it. I am on Mac OS Monterey 12.6.7

 

Last edited: Dec 17, 2023

try to do a full update to the new version 3.2.3.

try to go to. the Local Data file menu and then click on remove local sever. This will re download the local server.

 

@curtified Just wondering. Do you have an unofficial ETA on the plugin version of this awesomeness!!?

 

Roxette-Marie-Fredriksson-Necessary (Long Version)

Roxette-Marie-Fredriksson-Me And You And Everything In Between

Marie and Per Back together one more time Thanks to https://www.tryreplay.io/

 

We currently have one, but it's really buggy. I'm afraid of the backlash I'll receive if I release it without thorough testing. We're facing an issue with timestretching audio when the BPM is changed.

All the plugin will do is allow you to transfer the audio into Replay, much like Melodyne. You press a transfer button, play your DAW until your audio is in Replay, and then convert the audio to the voice or instrument of your choice. When you hit play, it will play the newly converted voice, and you have the option to drag it out of the plugin into your DAW.

Currently, I tend to drag audio from my DAW to Replay for conversion, and then drag the converted audio back from Replay to my DAW in standalone mode. It essentially accomplishes the same task, perhaps even faster than waiting for my timeline to transfer the entire audio into the plugin version of Replay.

With that said, we'll have a plugin available soon. The only difference is that it will open from inside your DAW and allow you to use additional plugins afterward, so you can hear your vocal chain in the context of the song without having to drag and drop the audio.

 

ROGER YOURE A BEAST!!!

 

Roxette - Feat - Marie Fredriksson Vocal AI - Sunflower

Roxette - Feat - Marie Fredriksson Vocal AI - Jezebel

Roxette - Feat - Marie Fredriksson Vocal AI - Walking on Air

Roxette - Feat - Marie Fredriksson Vocal AI - The Craziest Thing

7 songs have been completed from Marie Fredriksson Aka Roxette trained vocal dataset.
Marie was not available to record these songs with Per her band colleague due to her passing
Vocal AI Technology puts Marie back with Per in Roxette.

RVC2 Nvidia1600 was used locally to train Maries Vocals
Maries Trained vocal pack was then used used within www.tryreplay.io
weights packs for replay are Available free from from www.weights.gg

 

@Rodger So when can we expect the tour? Maybe an AI holographic show & residence at a Vegas hotel?

 

lmfao that shit's wild
tried American Idiot sung by Trump, pricelesss

ps: no politics, just fun

 

Its amazing what these AI technology tools
Are able to acheive
Little tip when creating trained packs do Not include vocal multi stacks example: leave out Chorus's in your datasets to train only use Sung Verse material vocal multi stacks will obsecure the trained weights with incorrect overtones.
No one would of ever thought this possible 10 years ago

 

Last edited: Dec 18, 2023

That would be very interesting to hear brother
Throw us all up a link

 

here's vocals only (to prevent copyright shizz on SC):

 

All of the tools are mentioned I this thread multiple times.

Suno X UVR X Replay. ETC

 

question/suggestion @curtified - what settings to tweak when result replaced vocal is too silent? it's most likely problem of a dataset, but even with settings adjusted, balance between replaced vocals and rest of the music is off, any ideas?

 

Yeah its most likely due to the data set. Sometimes I do the model merge feature and put another vocal at a low percentage that will help the main vocal come up in volume.

the other thing is turn up the stem separated vocal in your DAW and put it back into Replay

 

also one more question @curtified - when adjusting pitch of both the vocal and the music (for ex. +6 and -6, to preserve +12 octave of vocals against music), I noticed tons of artifacts and clipping - I assume workaround would be pitch shift whole song in a DAW first, and then do the standard +12 vocals (and +-0 music) adjustment?