curtified, is that YOU, or is it the app talking? Because I keep seeing a clone of these words in posts here and in other threads



Listen guy, whether or not this is for me has yet to be determined. Relax. I am merely trying to find a solution and get answers to a few very simple questions, for myself and many others who keep asking the same things over and over. Stop ignoring and dismissing us so easily, some of us are nice, if not slightly daft like yourself, people too!

 

You have to switch replay to multi-model mode. then choose your models and tick the merge feature. You can then pick the percentages of each model.

NOTE!!!: currently, the app only works with model versions that were created with the same version features. It might take some trial and error to find which models work well together.

 

As for the a clone of words. I am trying to complement and help everyone in all posts on this forum with our app. Im sorry if they are similar sounding solutions and feedback across posts. Thank you for the critique of my use of being responsive on this forum.

I am not ignoring or dismissing. I am actively working on solutions to fixing these problems.

This isn't my main job. I created this as a tool to help and when I'm free from other work I put my time and attention into this app and responding to this forum.

 

I think I speak for most people when I say, we applaud you for your efforts. The frustration I and I assume others who are genuinely interested in this are having, is the lack of clear instructions on how (and where) to install, download files and operate the app.

When I say my DAW computer is offline it means strictly offline, not a casual switch flip back and forth, online sometimes for this and offline sometimes for that. Saying it works offline isnt the whole story if you MUST be online to install the app and factory files for it to even work - then you can turn off the internet and work offline, right?

Ive yet to see an answer to whether or not you can provide a direct link to download the 18 essential files and if they can then be installed on an offline machine?

If not, then is it possible to install the app, open and download the 18 files, then bundle up and transfer the entire lot (app + 18 files) over to an offline machine and repeat the process?

Again, just trying to find a simple solution for an offline computer scenario!

-----------
To recap: (and please feel free to correct and/or use my tutorial since youre so busy right now)

1. Download and install the app.

2. Upon opening for the first time, you MUST download the "factory" models content (18 files 5-8 GB) which requires internet connection to do so. The app will not work without this content installed.

3. Download and install the "user" models you want to use.

4. NOW you can turn off that pesky internet connection and use the app offline (on the same machine).

*If your PC is strictly offline, you wont be able to complete installation. You MUST be connected to the internet to install the mandatory 18 factory model files required for operation.

 

Last edited by a moderator: Nov 12, 2023

I am (was) an acountless lurker for about 2 years now; Love this forum, love (some people) in here; I will abstain from any comments regarding how some people reacted to this dude just trying to help.

On point tho: am I the only one who instead of this remix idea sees a huge potential in the text to speech application? Is there any way to use this software for this? I've downloaded, had fun with some songs but, but the text to speech thing would actually convince me to pay for it. The interface and everything is super clear and organized; I used to use the old UberDuck, but they've pivoted from text to speech to something music related and rn are pretty much dead.

 

Thank you for consolidating everything from this thread into one post.

We are currently working on a way to have a fully "portable" version. Until then we will have a link to the files needed in an offline download this should be in the next update we push once we fix this list of bugs.

 

Welcome to the posting side of AS!

We have explored text-to-speech solutions for this purpose. However, what sets Replay apart is its speech-to-speech capability. This allows you to rap, sing, talk, or even use UberDuck audio as a reference audio. Then, Replay can alter the voice to any model from our extensive database of over 16k+ voices.

Replay works best with dry audio, with no vocal stacks, reverbs, or delays.

 

Simply AWESOME! Thank you for listening to us annoying AF offliner's

I really look forward to the update!

 

Its not annoying. Im a traveling DJ for a living I'm on flights most of the week so fully offline is why I built this app in the first place. The thing didn't factor in is the initial setup that needs to take place.

We will have that solution super soon!

 

Cool, this opens up so many creatives options, i was wondering if i create a song with my own melodic line and then i blend different models..
could that be used commercially?

 

I will try to record myself and play around lol; Also, besides some of my defects, I am a digital horder, and will probably pivot in the next year or two on full offline with my workstation; Is there any way to see where the models (downloaded via App) stores the files so i can make a backup copy? (On windows btw)

found it lol; nvm

 

That isn't my wheelhouse. Its a longer discussion that will be solved over time. Just like stable diffusion, midjourney, and dall-e being trained on Getty images. New laws will be established in the areas of how AI data is trained.

I would suggest not releasing your song as "Artist - My cool song feat. 30% Drake and 70% kanye". Try "Artist - My cool song" If your multi blended model sounds good and unique then you have a new sound!

or you could just be the next Led Zepplin clone like Greta Van Fleet

 

Last edited: Nov 13, 2023

All of our models are on our website that's in the orig post www.weights.gg feel free to digitally hoard all of them.

 

I appreciate your work. The problem that has created my reaction as "wasting time and resources" comes from the misleading definition of "cloning" voices in the introduction. And look at the examples that people have uploaded to that site. It seems valid for me that so many people deluded themselves trying to create samples sounding some famous person. It does not sound that person! No, you delude yourself if you think you will create a convincing Freddie Mercury voice. It sounds a pitch-shifted, many times granulated, over-processed sound.
Well, just like nowadays pop music. So, you can argue that it sounds legit for the current era, for some.
But this is stupid and will die one day. These over-compressed, over-autotuned shit will have an end some day... believe me. Just look at the Stranger Things effect on the new generation that found more interesting the sound of those old songs than listening to today's shit.

Now, there may be usefulness of this app. Someone commented on AZ (I don't know if it was you on the name of Sexyeyes02) some interesting use cases:

- Need a backup choir? use 5 different vocal models to create a dynamic backing choir to your original vocal takes.
- Are a male singer looking for female backup vocals? Use any of the female vocal models to create a nice harmony.
- Have a melody line you hummed? Try one of the instrument models to have a cloned instrument perform your hummed melody.

That would be awesome to hear what this app can do in these areas. Is it going to sound natural, not too over-processed?
So far the voice synths that exist (Vocaloid etc.) still sound fake in term of distinguishing male voice from female. Is it just due to the badly chosen range for that voice or the nature of the technology? I don't know.

I suggest you to use a better definition for the app than the word "cloning"!

 

Hi,
Since yesterday I've been trying to use the app but the stem step crashes every single time.

Am I the only one to face this error?
Without the stem step I am able use a model to convert voice and the result is pretty good.

I tried every combination in settings, most of stem methods and other settings with no luck

 

Your error is only involving the stem separation? Try a full reinstall and see if it fixes it..

Discord members are dealing with the same issues since we updated 1.9.1.. We are currently working on a fix for this.

 

Last edited: Nov 13, 2023

The user you mentioned from AZ is me. Those suggestions were just a few of the many potential uses for this app.

Another potential use for Replay is to create a high-quality model of your voice using a premium microphone combined with your favorite compressor and EQ. You can then process various vocal recordings, whether they're captured with an SM58, a voice note, or other low quality means. This will allow you to achieve studio-quality vocal takes for your songs, even from lower-quality reference recordings.

Your anger and projection suggest there might be deeper, personal issues to explore, as our reactions often mirror internal challenges.

Regarding music, your views are valid, but they seem influenced by confirmation bias about what constitutes 'real' music. New tools and technologies like compression, auto-tune, and AI don't negate traditional methods; they simply add new dimensions to the creative process, allowing old and new techniques to coexist and enrich music production.

By the effort you've put into this message, it seems you're concerned about the terminology we've chosen for our software. Could you suggest a more appropriate name for us to use in our promotions?

 

Last edited: Nov 13, 2023

I updated the google colab for training:
It should work now.

UPDATED LINK:
https://colab.research.google.com/drive/1DqRzmO-TO2z-DsJyzShfAysUvPxz-EdX?usp=sharing

 

Where can I find the MOVIE voice? This is impressive!

 

Personal issue = my personal opinion
The quality of (listening to) music depends heavily on the dynamics. Compression ruins it. Am I wrong?

I can't understand where you are trying to get with this sentence to. AI adding new dimensions? My concern is only about the quality, not the method.

Based on existing (pre-recorded) voices you can create new ones with the characteristics of the former.

Edit:
I actually have a better idea, instead of Replay, it is more appropriate to say "Repaint voices".
(I'm not criticizing the name of your app, it wasn't my complaint, I'm giving this suggestion to the definition/description of what it does)

 

Last edited: Nov 13, 2023