(Tutorial) Use YOUR Voice in AI Cover Songs with Replay and RVC
HTML-код
- Опубликовано: 18 май 2024
- How to Create AI Cover Songs with Your Own Voice Using REPLAY and RVC
Get Replay for free: www.tryreplay.io/
Install Pinokio to load RVC and many other AI programs easily: pinokio.computer
Explore hundreds of voice models: weights.gg/invite?inviteCode=...
NOTE: Please understand that AI program like this often come with some installation problems and compatibility issues. My ability to troubleshoot issues, is limited. I would also ask that you consider that just because you may have an issue with software like this, it doesn’t categorically mean that “it sucks,” which is what I see a lot of people say when they have technical issues.
That having been said:
Today, we return to Replay and respond to some of the most common comments we got on our last video.
We address installation issues, and show you how to get set up with RVC to create voice models of yourself or others to use with Replay and many other compatible programs.
The video covers the technical requirements, including the need for an Nvidia GPU, and provides a step-by-step guide on installing REPLAY and RVC, generating training data for a voice model, and using this model to create cover songs.
It also explains how to download voice models from weights.gg/invite?inviteCode=..., adjust song keys to match voice pitches, and troubleshoot common problems encountered during the process.
00:00 Introduction to AI Cover Songs with Replay and RVC
01:22 Exploring Replay: Installation and Common Issues
05:15 Downloading and Installing Voice Models
09:00 Creating Your Own Voice Model with RVC
11:16 Training Your Voice Model: A Step-by-Step Guide
16:07 Testing the Voice Model in Replay
18:59 Final Thoughts and Call to Action
👍 LIKE If you found this video valuable. 🙂
🥰 SHARE If you know someone who might enjoy this video.
⏬ DOWNLOAD or ADD This video to your PLAYLIST for easy access later.
💬 COMMENT Your thoughts and questions are welcome!
📝 SUBSCRIBE / @bobdoylemedia
That's what keeps me going!
🔔 And make sure to hit the NOTIFICATION BELL to stay updated! 🔔
🌎EXPLORE
bobdoylemedia.com
www.contentcommando.ai
www.contentcommando.ai/virtua...
🎁 GET FREE GIFT - CASE STUDY
Discover What Content Commando Can Train Your Team To Do!
www.contentcommando.ai/#_kys1...
📔 ABOUT BOB DOYLE
We live in a freaking crazy time, don’t we?
Artificial Intelligence has exploded into the mainstream, and reactions are varied, to say the least. But for me? I’m all on board! As a creative, today’s technology allows me to take the skillsets I’ve acquired over a lifetime and use them like never before. Instead of feeling threatened by how rapidly seems to be taking of people’s jobs, I see the technology as a way to foster creativity in far more people than are currently tapping into their creative potential. I believe we are here as humans to create. And for me, it’s not an option. If I’m not creating something, I’m dying inside. So Bob Doyle Media is a showcase of my various outlets of creativity, and you’ll see that AI plays a big role in it. From virtual photoshoots and art and animation creation, all the way to voice cloning and face replacement… I play with all of it, and this is where I share it. If you’d like to work together, please reach out. I’d love to connect and talk about how I can contribute to your project.
🗓️ MY HISTORY
🧠 meetbobdoyle.com
🤫 Featured Law Of Attraction and Neuroplasticity expert in the book and film, “The Secret”. • The Secret
🎙️ Voice Over Artist 30+ years
📻 Broadcaster / Actor / Creative
❓ASK ME
Got a Question? bobdoyle@me.com
📌 FOLLOW ME
Facebook: / bobdoylemedia
Instagram: / bobdoylemedia
LinkedIn: / bobdoyleai
🪙SUPPORT
If you want to support me, the best thing to do is to share the content… sharing is caring!
If you want to support me financially (completely optional and voluntary, but a lot of people have asked for this):
www.thebobdoyleshow.com/tip
Learn more from Bob Doyle Media: www.bobdoylemedia.com 
Развлечения
What a great tutorial, thanks for making this Bob! And a great exposé on some embarrassing bugs we need to fix... Looking forward to seeing what everyone creates!
- Weights & Replay Team
Thanks so much! I’d love to record an interview with someone over there to do an overview video with info straight from the source!
This is cool, I was looking for a way to train my voice so I could have a copy of the file and do what you're doing.
I’m anxious to see your content on Udio.
God bless you brother! This video is a lifesaver, all one could ever wish for in one video. You are the best! Kudos! ⭐️⭐️⭐️⭐️⭐️
I've been playing in the same way for quite a while, it's been really fun, and I even used it to have virtual duets with my friends (I call it "us from another parallel universe").
thank you kindly sir!!
@17:27 that was NOT too low; that was actually quite good!
When I listened back during editing, I agree with you. I could have stopped right there. :D
Wow really nice video tutorial , I really do need a new i9 desktop machine with the best GPU.
Great introduction to hybrid AI.... auto gen plus your own voice. FYI.... +/- 12 Semitones = 1 full Octave. I will need time to try this. In past, I have performed Acapella. My natural Voice is Baritone, but can hit parts in both Tenor and Bass. As I have grown older, my vocals have changed, now on lower end of Baritone.
Hey Bob! I found something super useful. You can actually sample the weights voices SINGING instead of speaking. If you click on the weight, it opens its page, and there are five or six options to test the voice singing, and by clicking on one of them, it triggers the main window to sing instead of speak on all of the voices 👍👍
Yes, there are a world of cool tools to explore over at weights.gg!
Cool video.
you the man
You sound real similar to our weather man. James Spann. Every one's favorite weather man.
1)you need nvidia for pinokio to work or theres orher alternative for thoss who doesnt have it?
2)how much of words/how long to say when i talk on microphone?
Thanks for this tutorial. After the model is trainined with the RVC software, I can't find the weights folder. I have D_50 pth, D_100 etc to G_50 pth etc etc but none of these pth files work after I import to Replay as they all come back as error. Any advice on this would help
Sorry, under the RVC folder look for the assets folder, inside that folder you will find a Weights folder and they should be there
Cool. I already used your earlier videos on RVC and Replay to do this last week. Have you found an ideal length of scripts for creating voice models? I read about 4 scripts at 2 1/2 mins each. I also trained for 500 epochs but not sure if there is a better return for way more training cycles?
I’ve never done that many. Up until this video I always used 20 minutes of audio but 10-12 seemed to work just fine.
Thanks for the great video! I tried to install Pinocchio on my 3090 system, but it wouldn't work. There are so many programs I want to use in Pinocchio. I hope I can figure it out. I don't' want to have to search out and install all of the programs separately.
I‘d like to point out that you neither need to extract the ZIP nor rename its contents. You can simply drag in the ZIP as-is and it use its filename (without the .zip) extension as name for the voice.
Yes, several have mentioned that. If I do another Replay video I will make a special point of that!
Help plz: I have run the insulation for the RVC now many times and the Zip, Git, & Conda will not install. Why not sure. But what do I do now plz?
I'm curious about using A.I. to fill out existing compositions... Background vocals... Exotic instruments... Pads, etc.... rather than singing the main vocal parts, the compositions and the main cords, solos, bass and drums....
At last, a really great walk-through for RVC on Pinokio! I’ve lost days to getting RVC to work properly like this, and I confess, I’d given up. Thanks Bob!
I hope it works for you! I know Pinokio can be problematic and very frustrating sometimes!
Does it support other languages' songs beside English? like Korean, Vietnamese, French, etc..
problem is the models en server download all data to the c disk, were windows is installed en soon runs out of space
P.S. I have an RTX 4090 and it looks like replay uses your CPU for track separation, and then the GPU for the vocal exchange and song creation. I was checking out task manager when I did this, and it hits the CPU to about 65% when track separation occurs , once that's done it hits the GPU to its Max with almost no CPU to change the vocals and create the new track. Just thought I would put that out there in case anyone is wondering.
thanks for the tutorial but all the voice clones i’ve tried sound rubbish like your results, very metallic overturned? and nothing like the person i was trying to clone
I installed pinokio and download all process. but it wotn show the main page like yours where it say RVC WEBUI..... what i can do? ?/:
Audacity now installes in a dir called Muse, ( C:\Program Files\Muse\Audacity) i have put both those files in there but when I open edit pref they are not there ? ( C:\Program Files\Muse\Audacity\openvino-models and C:\Program Files\Muse\Audacity\Audacity-OpenVINO-Module-3.4.2-R1) what now?
FYI, you can drop the .ZIP of the RVC model directly in the Replay model window. Replay will unzip and install it with the name of the downloaded zip file.
I’m so glad you said that. I was wondering about that, and I never got around to testing it. I really wish I could retroactively say that in the video because that makes things so much easier.
Came here to tell you the same after the last video you made =)
where does the latest version of Replay save the songs?
You mentioned all the pc requirements for the best results, what about M series Macs?
All M-Series Macs support Replay as well. Replay currently supports Mac M-Series, Mac Intel (with limited performance), Windows, and a Beta build for Linux.
Hi Bob! The program has been updated to v.4.1.0 It has added new features. I couldn't find a complete guide on how to use it. Do you know where I can find it?
I'm going to be doing an interview with someone to do that very soon. I was going to talk with someone from Replay, but they are recommending one of their power users, so we're just getting that set up. Stand by!
Ок
I wish for a cloud based Replay. Tried to install it but when downloading the AI model it totally ate my C-drive (+10GB), despite I had installed Replay on another drive because I don't have enough free space left on my C-drive. So I can't run it.
thats a real problem to put every data on c , now i have to delete replay for using all other programms on my computer
@@festivaltv Yep, I simply do not understand why the program doesn't allow you to choose which location the model should be downloaded to. Requiring free access to more than +10GB on the C drive for that is insane.
Weights is exactly this! Weights supports all the same functionality as Replay, but in the cloud for free. You can also train your own models and use them right there.
Hey can you explain where im meant to find the rvc voice file when its done loading the epochs?
if you cant find it, just search it in the file docs bruv, thats work for me
You for got to mention the UVE MDX NET VOC FT MODE
Hi Bob, thanks for your video, it was great but my cmd said that "The system cannot find the path specified." after i installed RVC in Pinokio, can you help me fix this please
Can you clarify at exactly what point your getting this error?
When i downloaded RVC in pinokio, i launched it and it started to run but the cmd/code on pinokio screen stopped at the line, report that : "ModuleNotFoundError: No module named 'dotenv'"
I don’t understand where you’re getting the music from Ken. I just download karaoke song.
Where does the pth save once its done training i cannot find it
same problem over here
Only thing you missed is where do the converted model files go? What is the file location?
They are in the “assets/weights” folder in the RVC folder. So sorry I wasn’t specific!
Is RVC your first choice for cloning a voice?
It’s the only tool I know that does as good a job and is open source.
wish that the models (big data space) set to another disk than c system disk
Thank yo,very interesting! But I have not understand where I find the PTH file when finish with the terminal... (sorry, u speak very fast for me, I am italian... :-)
You can just select the downloaded zip file directly, no need to extract it
👍👋🏻
My laptop does not have Ndivia, Does that mean I can’t run “Replay”? but it have a good processor.
Most AI related processing is done on the GPU, specifically Nvidia GPUs utilizing the massive parallel processing of these cards. Historically many people who were not using their laptop for gaming chose laptops without a dedicated graphics card or with one by Intel or amd, which work fine for most tasks, but sadly are fairly useless with AI. To train a model you will really need a GPU and if you haven't got one best bet is to use colab (a cloud computing system) there are tutorials on this. Once you have a model the grunt work to convert a voice to the model voice is less onerous so can I believe be done on capable non nvidia units, though likely the processing will take way longer.
You can technically run on CPU-only, but performance will be extremely limited.
guy changes camera angle says its a vocal booth
Replay isn't good enough to sound like the actual voice. Use Applio or RVC itself to make covers.
Reflection: Just replacing the vocals with another voice, does not remove the ethical aspects of using a track generated by AI to make money from it. 😁 The music productions that the generated music was "based on" is so much more than just the voice. The only thing added is the lyrics (that most people also AI generate). Think of the prompting as a filter that narrows down the output from an endless database of tracks. That will ground your creative "euphoria" for what you actually have made. That is actually similar to finding a hidden track on a streaming site, that you kind of "claim as your own".
At least you managed to demonstrate to us how awful and low fidelity the SHURE dynamic shit mic is that you use.
Even most $100 condenser mics beat that overpriced turd.
You weren’t listening to the audio from the shire. You were listening to the audio from a cell phone. The audio from the shure went into the model.
I meant to clarify that.
I was cloning the voice using Tensorboard, but during processing and training, I kept encountering this message frequently:
"FileNotFoundError: [Errno 2] No such file or directory: 'assets/rmvpe/rmvpe.pt'"
"f0fail-96-C:\Users\Ryzen\pinokio\api\SA\app/logs/untitled/1_16k_wavs/0_33.wav-Traceback (most recent call last):"
Unfortunately, I personally don't have any experience with tensorboard, but maybe someone else can help. :)