How To Train AI Voice Models ONLINE For FREE (No GPU Needed)

[@lettucepicker]

People who are complaining about pretrained model error. Just run "Download Pretrained Model" without changing anything after running "Set Training Variables" 4:54

[@postmanager5604]

hello i have a question, after everything done i cant find go-web.bat file, whats the problem? do you maybe know?

[@StijnToonen-jj4tw]

Thank you for your reaction, it works, Thanks to you

[@quas3728]

Bro this worked like a charm. I tried 5 minutes voice conversion with 50 epochs like your video. the quality is nice. thank you again!

[@theAIsearch]

nice!

[@navierproductions709]

Bro it shows " No features exist for this model yet. Did you run Feature Extraction?" tho I have run the preprocessing and feature extraction process.. pls help me with this issue thanks alot

[@KaleshiBabaUP]

@@navierproductions709 hey in my case while traning for the first time its showing :You need to download a pretrain! Please run the "Download Pretrained Model" cell before continuing. What to do to fix thhis ?

[@KaleshiBabaUP]

Same in my case

[@TheRapist4youu]

The only youtube channel which dosent adds weird confusing coding things. The way you explain is fabulous. I have never seen someone on KZbin explaining hard things like this simple

[@theAIsearch]

Thank you so much!

[@amykaushik5998]

​@@theAIsearchI need your help

[@navierproductions709]

@@theAIsearch Bro it shows " No features exist for this model yet. Did you run Feature Extraction?" tho I have run the preprocessing and feature extraction process.. pls help me with this issue thanks alot

[@navierproductions709]

Bro it shows " No features exist for this model yet. Did you run Feature Extraction?" tho I have run the preprocessing and feature extraction process.. pls help me with this issue thanks alot

[@urasam2]

@@navierproductions709 Have you tried putting the zip folder into your Google Drive directly, rather than into the folder in the interface? That worked for me

[@IIPresidenti]

I have to give credits to you, my guy, without you my channel wouldn't exist

[@chaozowl]

DAMN

[@MeganeJourney]

I've been searching and watch some videos on KZbin. So many errors and fails because I just use Radeon RX580, and then I find this awesome video ! I watch this tutorial step by step and... Now I have my own voice models just by using cheap and old GPU. Thank you very much ! You are amazing. Greetings from Indonesia.

[@theAIsearch]

Glad I could help

[@swiftuihok]

wailah bang ngapain sih buat apa sih bang wkwkw.

[@a-m7004]

Last time my computer caught fire because of local RVC Thanks for this video❤

[@theAIsearch]

My pleasure. Hope your pc is ok

[@Gael-k7q]

oh thank you so much, i've been struggling to make this work for me, i had so many doubts and you've cleared absolutely all them in this video, really thx man!

[@theAIsearch]

Glad it helped!

[@YoYofan17]

this was really helpful bro compared to your past videos which you have uploaded... thankyou!

[@navierproductions709]

Bro it shows " No features exist for this model yet. Did you run Feature Extraction?" tho I have run the preprocessing and feature extraction process.. pls help me with this issue thanks alot

[@urasam2]

Thank you so much for this. You made a really clear tutorial, and the software ACTUALLY WORKS!! I was using another notebook which was working OK until recently, it's now useless. Tried this one this morning and it works like a dream. Only thing was I couldn't get it to work by placing the zip file into the folder in the programme but the other method (putting it into my drive) works just fine.

[@theAIsearch]

You're welcome!

[@FullOfDreams-lr1wf]

No way I actually understood how to do this 😭 this guy's great at making complicated things look simple

[@theAIsearch]

thanks!

[@Duegobrasil]

Bro please reply me 😭 I followed step by step but when I downloaded pth file from my drive then what to do next??? I mean how do you get mangiorvc file and goweb. Bat file??? Help me there are no files in my PC folder? 😭🙏🙏🙏🙏🙏🙏🙏🙏​@@theAIsearch

[@cargadorcompactador]

hello ​@@Duegobrasil

[@theAIsearch]

put the .pth in a zip. then upload to google drive. then use the drive link to 'download model' here: kzbin.info/www/bejne/ioSvfmSjeqaWZ5o

[@personagenscantandoraps]

1:11 in fact, you can train a model locally without a GPU, you can use your CPU, the only problem is that this will be very slowly, but you can!

[@connorblacknight6272]

omg really? how?

[@alicebyno]

Woooww, thanks so much for making this noob-proof. I only started looking into voice models a couple of days ago and today I made my own thanks to you 🙏🏻🙏🏻

[@kevinrycmog]

wait so that isnt ur real voice.

[@st0v0n]

Under rated comment 😂

[@xyerx]

discovering your channel felt like Columbus discovering America

[@bensmith8463]

I have been using all your RVC tutorials and they have been immensely helpful for me, thank you! If I wanted to resume training offline on a model that I initially started in Google Colab, would that be possible? If so, which files are important to copy over? Do I only need the G and D files?

[@panosjr_greece]

The most powerful channel in KZbin. Well-done mate. Can i use other language with google collab?

[@theAIsearch]

thanks. yes, any language would work

[@MaxSparky]

Your Unreal Absolute Legend!, Straight forward and easy to follow

[@theAIsearch]

thanks

[@navierproductions709]

Bro it shows " No features exist for this model yet. Did you run Feature Extraction?" tho I have run the preprocessing and feature extraction process.. pls help me with this issue thanks alot

[@thomasforthsen8360]

amazing!!!!! been looking for this type of genAI stuff for a long time. Thanks

[@theAIsearch]

Enjoy!

[@navierproductions709]

Bro it shows " No features exist for this model yet. Did you run Feature Extraction?" tho I have run the preprocessing and feature extraction process.. pls help me with this issue thanks alot

[@LavenderHarper-n3u]

@@navierproductions709 im not sure if this will help you or not but this happened to me when i only had one audio file in the zip folder instead of two or more so what i did was just add another audio file to the zip file and repeat all the steps from when you add the zip file to the folder in google drive

[@BrokenMirror-dh9xj]

Great video man, but I have a question. How do I properly resume training on my models? Like, do I have to preprocess my dataset again using the cell, or should i just use the "load the preprocessed dataset (to resume training)", cell?

[@JuanTheGirlAffinty]

Hey Ai Search, I had few questions to ask you 1. How many epochs is the best we can do with it 2. What happens if you had an like an 20 minutes sample or 30 minutes sample of the voice? 3. What is the code to use it to prevent google colab for to detect it for inactivity? 4. What is the limit usage for when you are using RVC v2 Disconnected? 5. How long does it take for our connection disconnected when using RVC v2 Disconnected? Well that’s it for my questions

[@bdekid6499]

At least one Guy that Speak my Language VIELEN DANK Thank You

[@SPC_TheFurryFriend]

AI Search, you are my hero! *blows a kiss*

[@alessiobocci7127]

No GPU detected; training cannot continue. Please change your runtime type to a GPU.

[@yosh839]

This is such a great tutorial! I'm super new to all of this, so these might be dumb questions, but if I were to train my own model, would I then be able to upload it to a voice changer program to use in real-time? Also, does RVC have a recording time/upload limit? I've been wanting to try out content creation, but my anonymity is really important to me, so I'm hoping to train a voice model with my own voice (of course, I'd be doing a voice that doesn't sound like myself) then use that voice for content creation. My main thing would probably be KZbin, and I'd likely be doing long form content, so I've been looking everywhere for a voice changer that would be able to handle like 30+ min of audio or recording because doing a unique voice for 30+ min on my own every time would kill my throat. Do you think RVC would be good for this?

[@theAIsearch]

yes, you can upload the pth file to use in realtime in w-okada (kzbin.info/www/bejne/lXu7k6mJiayDpdE) this huggingface space probably has a limit, but I haven't maxed it out yet. if you do, you can also install rvc to your pc to run it without limits. see this kzbin.info/www/bejne/n6mlaqKXoblmmbM rvc is def the best & free ai voice changer out there, so you're on the right track!

[@yosh839]

@@theAIsearch Awesome, thanks so much! Appreciate the additional video links as well :) I'll be checking them out for sure!

[@postmanager5604]

Is it possible to upload pth file to huggingface and then basically use text to speech generator?

[@terranaryder583]

Thanks so much for this amazing tutorial! This explained everything in such a clear and easy way, even for somebody as tech illiterate as myself, haha. If you had time, I had some questions I'd like some help with: 1. I left total_epochs on the default setting, 500. I have about 15 minutes of voice data for my character. Is 500 epochs an appropriate amount for how much data I have, or is that overkill? 2. Once my model reached about 340 epochs in the training, Google Colab disconnected itself because I'd apparently reached their GPU limits. When I tried resuming today, after running the dependencies and setting the training variables to the same as last time, I just skipped to "Load preprocessed dataset files from Google Drive" and input the step count as 2333333. Just want to be sure those are the correct steps for resuming? 😅 3. When I restarted training, the epoch count in the log started back at epoch 1. Does this mean my progress from my previous session was lost, or is it normal for the epoch count to reset with each session? 4. I noticed in the log, the percentage number after every epoch is 0% (see example below). Does this mean I did something wrong, or is this normal? INFO:a_experiment:====> Epoch: 52 [2024-02-24 23:44:31] | (0:00:25.563601) INFO:a_experiment:Train Epoch: 53 [0%] INFO:a_experiment:====> Epoch: 53 [2024-02-24 23:44:56] | (0:00:24.366461) INFO:a_experiment:Train Epoch: 54 [0%]

[@theAIsearch]

500 is too much to start. try with 100 or 150 first. You may hit the daily free limit after 340 epochs so try 100 first and let me know how it goes. the 0% is normal. that's the start of the epoch. the next line shows its completion

[@terranaryder583]

@@theAIsearchThanks for getting back to me! I managed to successfully finish my voice model, and it turned out perfectly! Though in my two training sessions combined, I think I accidentally ended up training it with around 600 epochs...at least the model sounds pretty good, haha. 😅 Some info for anybody else reading this, if it's helpful: - Definitely don't use the default setting of 500 epochs, or you'll hit the GPU limit halfway through. I personally hit the GPU limit at around 340 epochs. If you hit the limit, Colab won't let you connect to a GPU again for ~24 hours. - Can confirm that if the training gets interrupted, you resume by running the dependencies again, setting the training variables to the same as in the previous session, then skipping down to initiate "Load preprocessed dataset files from Google Drive" and inputting the step count as 2333333. - Colab doesn't remember how many epochs you trained in your previous training session before disconnecting. So let's say for example, you disconnected at 100 out of 200 epochs in your first session. When you resume again later, you should set the total_epochs as the remaining number of epochs you need (100), rather than setting it as 200 again. (Learned this the hard way...)

[@GeneralDonato]

@@terranaryder583 So to sum it up, let's say I want to train a voice model using 250 epochs, but because of the concerns regarding daily GPU limit, I decided to train it at only 100 epochs instead, and it succeeded. However, I still want to train it further to reach my original objective of having a 250 epochs voice model. So when I resume my training with the G and D files and the step count stuff, the only thing I need to do is to set the total_epochs to 150 instead of 250, and when it finishes I will have a 250 epochs voice model, is that what you're saying? Also, throughout the training sessions, because I extracted about 90-95% of a character's dialogues and then used all of them up in one single WAV file consisting of 28 minutes and 36 seconds training datasets, so I didn't have any new audio files. With that in mind, would the resume training part still works and improves my voice model even further? Please answer this as soon as possible, thank you.

[@terranaryder583]

@@GeneralDonato Hello! I wouldn't exactly consider myself an expert, so I might not be the best person to ping, ahaha. 😅 But I'll try to answer your questions to the best of my ability, based on my own experiences from training my model! 1. If I understand your question correctly, for the total_epochs value, you should actually put the total number of all epochs for all training sessions, not just the number of epochs you want to train in the current session. For example, if your model currently has 100 epochs, and you wanted to train 100 more epochs in your second session, you would put 200 for total_epochs! 2. At some point of the training, there will be diminishing returns, but generally, training more should still improve your model further, even if you haven't found any new audio files. However, there actually is such a thing as training too much, which might make the voice model more inflexible when given audio files that are too "unusual" or "different" from its training data. I believe this is called overfitting! I vaguely recall there being a few videos on KZbin about how to avoid this, though I can't recall them at the top of my head. Hope this helps a bit! And apologies if anything I say turns out to be inaccurate - as I mentioned, I'm by no means an expert!

[@GeneralDonato]

@@terranaryder583 I see. Regarding my first question, the reason I asked about the total_epochs stuff is that no matter what I do, I can never get the total_epochs I desire. I wanted my voice model to get to at least 600 epochs considering that my training dataset is 28 minutes and 36 seconds. The first attempt I actually managed to get to 200 epochs before the Google Colab notebook disconnected me for excessing their daily GPU limit, the second time (I decided to go for only 150 epochs since) I got disconnected at 50 epochs, and the 3rd time which I attempted yesterday evening, I actually managed to get to 149 out of 150 epochs. I was so close to finishing my voice model, but then I got disconnected for the issue with daily GPU limit AGAIN. It seems like my laptop at best can only get me to 200 epochs each training, but I still wanted my voice model to get to 600 epochs after numerous trainings, so I'm not sure what to do now. Either way, thanks for your answer.

[@vibeever]

it is very neat and clear .. thank you!

[@theAIsearch]

You are welcome!

[@fuckyoutubeyou]

its not clear at all. mine says to download a pretrain model from the cell befor start training. doesnt make sense he never even has this option in the video

[@idireel7765]

QDAMNNNN AFTER SEVERAL DAYS FINALLY THANK U MAN

[@theAIsearch]

Glad you got it to work

[@idireel7765]

@@theAIsearch ye ahah ty a lot !

[@RedCloudServices]

thank you for the video what is the process for creating an AI cloned singer ? would you just sing your reference audio and train again? or can you use the same spoken cloned voice

[@nbase2652]

I was wondering if there is any speed difference between training remotely (Colab) and locally (8GB RTX 3060-Ti, i5-6600, 48 GB RAM), so I tested it: Exactly the same dataset of around 50min manually split up into round about 500 files. I was using the exact same settings (batch size 8, TITAN pretrained base) for both the remote Google Colab as well as in my local Applio installation. Time it took for 1 epoch was around 1m35s (colab's default T4) and 0m55s (RTX 3060-Ti) respectively. For a 400 epoch model, that would be roughly 400min (RTX) vs 600min (Colab). Adding the time it takes for colab to start up (running installation cells at every new session) and all the hassle with timeouts, capped session time for free accounts (click the RAM/disk panel on the top) etc, I will definitely stick to training on my own machine. Soo TLDR; Colab is roughly 50-60% slower than my 3060-Ti, and Colab is both inconvenient and less dependable. Just thought I'd share this in case anyone was wondering the same thing. Anyway, thank you for the tutorial!

[@DealazonDaily]

As usual The Great without Any lag in video ❤❤ thanks Also Other then Google colab is there anything we can use in online for an free account that would be more useful if you have suggested something please ❤ thanks 😊

[@theAIsearch]

do you mean for training a voice, or just RVC?

[@DealazonDaily]

@@theAIsearch no for all like stable diffusion and fooocus like that I need cannot run in my local so is there any alternative like Google colab for free is available there like I am asking

[@nahhdappaul4442]

Thank you so much. for some reasons. the training process failed and I do not know how to fix it. At the training stage

[@joelrose8514]

Any idea if polyphonic instrument conversion is on the horizon in the near future? I used Mangio RVC to successfully convert a pedal steel to a resonator sound today. Although it was monophonic, it was still very surprising how good it sounded.

[@HorribleHomeVideo]

9 months ago you could do a few easy steps and in mins you had a voice cloned. now its all these hoops to jump through and i would rather pay google collab to go back to how easy it was

[@nehalamkane7367]

can you tell me which rvc model you downloded and from where it downloaded

[@GhulamMustafa-pm7wd]

same question

[@ArmaanHasib2.O]

bro Preprocessing and Feature Extraction results shwoing 'no-feature-todo' please help

[@leonsousa5004]

Same here, please help!

[@Sperm0L_Music]

same problem

[@tigrovica8417]

I'd like to add, that the audio files inside of zip file also should not contain any space or special characters.

[@theAIsearch]

thanks for sharing

[@grfdgfxd]

i had added a special character and still worked

[@savanpatel9198]

How do I download RVC? I tried to download Mangio RVC Fork from online. When I go to click the go web batch file, the command prompt opens and tells me to press any key. When I press any key, the window closes and nothing happens. What am I doing wrong?

[@nehalamkane7367]

then can you download rvc and from where and is it working

[@flotsy3177]

i tried both GPU and no GPU usage things (from ur vids) and everything runs clear untill idex training, it shows No features exist for this model yet. Did you run Feature Extraction?, even tho i did, for the one with GPU usage also some kind of error pops up

[@forfanw1ososbie990]

Run preprocessing and feature extraction cell

[@CarloIñigoJulag-ay]

same problem

[@lavendermilkhoney6743]

Had the same issue and fixed it, I just posted a comment about it if it can help.

[@lavendermilkhoney6743]

@@CarloIñigoJulag-ay Had the same issue and fixed it, I just posted a comment about it if it can help.

[@TrueCrimee1]

Hello brother, it was a good training, but I have encountered a problem. I trained one voice and set its repetition to 170 times, and there was no problem and it was correct, but I trained another voice and set its repetition to 500 times, but then‌ It stopped repeating 300 times and I clicked on the page once every 4 seconds. Thank you for answering me.

[@TrueCrimee1]

And I get the error (continued without GPU).

[@Prod.Dzl971]

(Error, no features exist for this model yet. Did you run feature extraction? ) However I extracted but it doesn't work. Can you help me ?

[@danzigz]

you probably made ur dataset a rar instead of compressing it

[@lavendermilkhoney6743]

Had the same issue and fixed it, I just posted a comment about it if it can help.

[@Holidayxx]

i have a question, i hope you notice my comment. when resuming a training with a new dataset. do you make a new zip file dataset? or you just need to add the new wav file in the existing dataset? about the new dataset, should it be renamed differently? or should it stay the same name as the previous one? when resuming training with new datasets, should i run preprocessing again? should i ignore it and run the ones that says (fur resuming) instead? thank you!!!

[@nehalamkane7367]

Plzz solve my problem index training problem cell not run i don't know what to do

[@leonsousa5004]

Same here

[@SSNintendo]

THANK YOU! You need more fame bro 😁

[@theAIsearch]

Thanks! I hope so

[@xboomie590]

Hi, so I wanted to add more epochs to my AI, but they are all stuck at 0%. I have even tried to start from scratch, and it's still 0% on every epoch. How do I fix that issue?

[@iwongmistis4592]

Thank you , this is answered of my request , Great job

[@theAIsearch]

My pleasure!

[@dollarproduction24]

Thank you so much, dear! It works❤❤

[@theAIsearch]

You're welcome!

[@leonardosales526]

Thanks! Now I will be able to train models again.

[@theAIsearch]

good luck!

[@arkadym8635]

you're a lifesaver

[@kinopiko01]

When I used Index Training, this pops up, although I have done feature extraction: Exception: No features exist for this model yet. Did you run Feature Extraction? Is there something wrong that I'm not seeing here?

[@danzigz]

you probably made ur dataset a rar instead of compressing it

[@TwigAndMushy]

i have the same issue, did you find a solution?

[@kinopiko01]

@@danzigz It's already compressed tho, in wav formats and all.

[@kinopiko01]

@@TwigAndMushy If I do find one I'll let you know asap

[@kinopiko01]

@@danzigz Exception Traceback (most recent call last) in () 26 listdir_res = list(os.listdir(feature_dir)) 27 if len(listdir_res) == 0: ---> 28 raise Exception("No features exist for this model yet. Did you run Feature Extraction?") 29 30 try: Exception: No features exist for this model yet. Did you run Feature Extraction? This is happening even though the dataset is already compressed into a zip folder, and feature extraction is already ran

[@thoseepiccommentsmedia5488]

Subscribed you're a life saver

[@star_shin_experiment]

you saved me. thanks a lot bro

[@Tontonmarto]

Hi! Subscribed to your channel. Thank you for this. I'm using the W-Okada voice changer and works very well with .pth file. But because I would like create French voices, english models voices have issue with the french "R" , "ON" , "EN" , "IN" pronociation. So I have no choice to create my own french voices models. I know some francophone actors wich can help me . The second problem I encountered is that the voice changer can't handle laughing and screaming. Do you know how to fix that ? Thank you again.

[@theAIsearch]

thanks for the sub! that's expected. you'll have to train the model with laughs & screams & other extreme stuff to give it more range.

[@AliasWasTaken]

This video is the perfect timing for me, thanks a lot for sharing this collab

[@theAIsearch]

my pleasure!

[@MJFANFR]

Hi thank you for your video but I have a problem my session is timeout but I don't have use colab more than 12 Hours do you have a solution ? already tahnk you for your video ♥

[@theAIsearch]

how many epochs did you set

[@MJFANFR]

@@theAIsearch 250

[@MJFANFR]

250 Epochs@@theAIsearch

[@MJFANFR]

​@@theAIsearchI have set 250 Epochs

[@MJFANFR]

​@@theAIsearch I have set 250 epochs

[@Dromitos1]

best tutorials on youtube

[@aliez404]

This is actually amazing. Holy sht!

[@noaharkadedelgado]

did it work for you?

[@alien10.3.05]

It still works?

[@noaharkadedelgado]

@@alien10.3.05 it worked for me but when I go back to update my vocal model it’s now saying I gotta pay

[@MatthewLewisMJ]

You are AMAZING MY FRIEND!

[@theAIsearch]

Thanks!

[@fakeplease1373]

Have a question, does more audio length give more accurate result Can i train a wav file with 30 minutes long? Will it take too much time?

[@theAIsearch]

That's correct, but usually 5-10min is good enough

[@fakeplease1373]

@@theAIsearch thanks man

[@twxzst]

how long does it take to do it again cause after i do it once its fine but when i do it the second time i dont see the rvcdisconnected or anything related to that?

[@_ueij]

Omfg. A good fucking tutorial.

[@BowGiulius]

Thanks alot for this one!

[@theAIsearch]

My pleasure!

[@domedynamo]

Plz plz in 17:00 how did you run the mango rvc web which is located in your rvc folders in file explorer well i dont have that bat file how do i run

[@AudreenAldeva]

i have an issue - i trained my model in 500 epoch - i get GPU limit in google colab which is the training is force stopped - i only have logs with those file format : seperated audio.wav, seperated audio.wav.npy, and index file how can i continue my dataset training with index file?

[@AudreenAldeva]

oh well for anyone had this issue, i fix it (probably?) i just use my second account to start over while watching the google colab gpu memory dont start with 500 epoch, start with 200 - 300 epoch

[@ell500]

@@AudreenAldeva i did do 200 epoch but it still disconnected me and i don't know how to continue

[@ell500]

@@AudreenAldeva i did 100 this time and now it still disconnecting me for max session time or inactivity even though i was not inactive

[@AudreenAldeva]

@@ell500 i use Auto Clicker to prevent me from inactiive

[@ell500]

@@AudreenAldeva i used a auto clicker too and it didnt work

[@SuperSonicBoy2001]

I got an error just now saying: "Cannot connect to GPU backend You cannot currently connect to a GPU due to usage limits in Colab. Learn more To get more access to GPUs, consider purchasing Colab compute units with Pay As You Go." This happens when I try to run the dependencies. Any idea what this means?

[@Seedley]

Can you take the model from the notebook and use it elsewhere like an API? Or does it have to be with RVC?

[@khanh1242]

Can you make more of video like this with more details and perhap other RVC model training sir?

[@LanNguyen-bv9wd]

Thank you very much! It worked pretty well for me ~

[@theAIsearch]

You're welcome!

[@poseidon_gaulz]

Because of this, my sis gone insane because thinking converstation with virtual waifus.... Great Job, sir!!

[@theAIsearch]

lol

[@poseidon_gaulz]

@@theAIsearch uhh... Ok

[@Mariolinosify]

No GPU Needed..... you need to have an Nvidia Cuda GPU otherwise you can't train your own AI voice models. makes sense

[@aggdub]

Thank you so much man 🥳

[@theAIsearch]

No problem 👍

[@oqualcycle]

How is the trained PTH model generated by "RVC v2 Disconnected" is used to then convert text to speech in your own voice? What RVC software does that?

[@fakeplease1373]

Waiting for the next video❤

[@khaioberlin]

for some reason, when i try to train it just says, "NameError: name 'pretrain_type' is not defined"

[@drownlore]

same, but I fixed it by scrolling up and downloading "Pretrained Model", which is above of the preprocessing part: this worked for me

[@exoenthusiast]

YOU ARE SUPPOSED TO SCORLL UP AND RUN THE PRETRAIN CELL TOO !! happened to me aswell

[@ItisMarlin]

At around 260 epochs while I was training at 500, the gpu connection was disconnected and failed due to the colab usage limit. Can downloadable trained models (epochs over 500-1000) be made only if the creator subscribes to colab pro?

[@bubyoz]

I want to train 1100 epoch xd any answer you found?

[@BazzilioSPb]

AssertionError: Hold up, you need to download a pretrain! Please run the "Download Pretrained Model" cell before continuing. What i need to do? I made all steb by step.

[@AgussxVenus]

i wonder if somebody could make a VocalRemover5 Disconnected since a lot of people (like myself) can't run the Ultimate Vocal Remover software

[@thomasforthsen8360]

hello again. i tried to run the codes and all went well except when i got to the training part, after hitting execute i got this error and don't know what to do. Could you help? error: NameError Traceback (most recent call last) in () 25 26 pretrained_base = "pretrained/" if model_architecture == "v1" else "pretrained_v2/" ---> 27 unpt = f"_{pretrain_type}" if pretrain_type!="original" else "" 28 29 pretrainedD = f"{pretrained_base}f0D{target_sample_rate}{unpt}.pth" NameError: name 'pretrain_type' is not defined3#for4 thanks

[@allchicken]

Same here

[@allchicken]

Well, looks like we forget to run one group that says "Download Pretrained Model", which is above of the Preprocessing part Maybe we didn't saw it because it wasn't on the video

[@garrone2000]

@@allchicken thx for the advice

[@postmanager5604]

Is it possible to upload pth file to huggingface and then basically use text to speech generator to get cloned voice?

[@XKathXgames]

I hope this helps me, my vocal chords are damaged from a fire, but I have a bunch of audios from my English classes. Hopefully this will help me out.

[@ElectronicBoxOfficial]

thanks for all your amazing videos and tut a quick question: any idea why i do always get [0%] on each Epoch? INFO:mali:Train Epoch: 47 [0%]

[@theAIsearch]

that probably indicates the start of each epoch

[@WallaceRobertoc1]

Obrigado por compartilhar seu conhecimento!

[@nobafan7515]

What should i set it to for training if i have an hour of training? Do i keep it at 8 audio files, or can i increase it?

[@JoeBacon2601Warrior]

Is there anything that can make the voice model to train faster in RVC v2 Disconnected?

[@Julius8888TW]

Hi, I think this video is excellent, but I have a question, can I use RVC with Chinese?

[@theAIsearch]

Yes you can!

[@Julius8888TW]

@@theAIsearch OK, thanks

[@ecotler00]

For me, I was able to fix features issue by making sure the dataset (the zip file) was the same name as the experiment name.

[@TahaLassiwala]

You kinda sound like J-perm. Anyways, great video.

[@swagatamsanyal6156]

this is the error when it run the training step - AssertionError Traceback (most recent call last) in () 7 8 ----> 9 assert 'pretrain_type' in locals(), "Hold up, you need to download a pretrain! Please run the \"Download Pretrained Model\" cell before continuing." 10 11 AssertionError: Hold up, you need to download a pretrain! Please run the "Download Pretrained Model" cell before continuing.

[@swagatamsanyal6156]

There is a way around - there is a step before which you add the zip files, the one that says run original model, you need to run that step before you run the traning model step, then it solves the issue.

[@shaka_can]

Where do we put the index file?

[@Molandria]

Are we able to use this to do speech, to speech during a live stream like on twitch? I'm trying to find a voice changer for that, which i can make a custom cloned voice for. But everything i look for, always shows me Text to Speech, or Recorded audio speech to speech. Never in real time.

[@osamaelawad1428]

Hi, have please update of same url google colab? Or how fix Not working, no access of drive,

[@Fred2817-ro8hf]

Is it possible to end the training early and use the model with the amount completed? I entered 300 epochs and it is taking way longer than I anticipated lol

[@xxxxRouvenxxxx]

Hi. Im using pretty good quality for the training but the endresult sounds robotic. What can I do?

[@weirdcollector]

how to calculate (or at least to estimate )the batch file size in relation to the total duration of the dataset?

[@killll]

worked perfectly, amazing! 10/10

[@theAIsearch]

Great!

[@shdwmere]

bro, are you the Fireship owner?

[@dollarproduction24]

Can you please show us how/where/what to write the code to provide info that we are active to google collab?

[@RyanRonin]

yes please!

[@Ziby3]

so what do i need to ask chat gpt for the code? 3:19

[@ahmedchawki8083]

i feel like god sent u just so i dont waste my money on a gpu that i dont need god bless u

[@theAIsearch]

my pleasure!

[@IDOLSKPOP68]

Hello. Can this train model be used for Tortoise TTS? Is there a way to train a model for Tortoise TTS using the same method as yours?

[@dimasky8671]

Exception Traceback (most recent call last) in () 29 listdir_res = list(os.listdir(feature_dir)) 30 if len(listdir_res) == 0: ---> 31 raise Exception("No features exist for this model yet. Did you run Feature Extraction?") 32 33 try: Exception: No features exist for this model yet. Did you run Feature Extraction? I ran the Feature Extraction though but the index step still doesn't work help

[@Evgeni25]

don't rename the folder inside zip file, when it already exist. create new folder on pc, rename it and drop it into empty zip, then drop audio into the folder. I could not understand what the reason was, all the audio was working properly. I recreated the zip several times and only then realized that this could be the reason that extraction not working, lol when the extraction is working, a lot of text with the percentage of work completed will be visible