Skip to playerSkip to main content
Training an AI voice model takes a long time, and you don't always have to do it in one sitting. In this tutorial, I demonstrate how to pick up right where you left off by resuming training of an RVC model in the Mangio-RVC fork.

I cover how to correctly load your previous experiment name, adjust your epoch settings to continue your training progression, and how to verify your results after hitting your target epoch count.

Original YouTube Tutorial: https://youtu.be/b0_m8kOVBXs
How to install Mangio-RVC: https://www.dailymotion.com/video/xa71r62
How to Start Training RVC in Mangio-RVC: https://www.dailymotion.com/video/xa71s8q

Video Details:
* Original Publish Date: January 3, 2024
* Focus: RVC Workflow / Model Training Continuation
* Test Using GTX 1660 Super

Follow lordcaocao2025 on Dailymotion for more technical AI research and generative workflow guides!

---
Connect with me:
📺 YouTube: https://www.youtube.com/@CaoCao2025
📱 TikTok: https://www.tiktok.com/@caocao20250
💎 Patreon: https://www.patreon.com/cw/Caocao2025

#RVC #AITraining #VoiceCloning #MangioRVC #AITutorial #lordcaocao2025
Transcript
00:00Hi guys welcome back with me Chow Chow 2025 this time we're going to learn
00:05about how to continue training after we train a models so go to train and here
00:13you need to know or remember the name of your current model going to continue
00:19with Yusuke Kitagawa here name of the model is Yusuke Kitagawa and
00:28setting up this like the usual and then you need to increase the epochs last
00:35time we train until 20 we could increase it like 200 or even 600 make sure you
00:42have set frequency like I put in 10 here and then simply click train models okay
00:49and it's gonna process it and don't worry if you somehow turn off the training
00:56even though it's not reached the 600 models it's going to continue again from
01:01the last frequency so we're going to do this going to train the models here it's
01:08kind of processing it read the last thing
01:17I'm going to continue the models it's gonna take a while before it started it's it's
01:24already epoch 75 so don't worry let's check point of seven 75 last time I did this so
01:32we're going to skip it and meet you at least at the eight t5 okay guys this is almost 22
01:41hours after the last time we play this yeah we reach for 82 epochs I don't think I
01:50could wait anymore so yeah we're going to end to 480 80 epochs because I save every
01:58time so if you feel tired you just want to stop it just just close this thing let's
02:05close this we're going to restart the ongeo okay guys I already restart the ongeo and test it out first
02:20you need to refresh this and then we're going to search the new one it create all
02:26the epoch again you could delete until the last one if you want we have this and we're going to
02:33use our
02:33sound sample that is this okay that's what we're going to use don't take a while
02:49cause it's loading oh it's already finished sorry okay guys welcome back to me load check out 1925 and
02:59last let me add so today we're going to learn about sparkline sparkline is some kind of graph but
03:06instead of making like instant graph square it put the graph in a cell so for example let's see here
03:16so insert and this is the sparkline that we're talking about okay you can make like five column and
03:23bin and losses line and sparklines for column to the right okay I'm trying to mask it around
03:33okay let's learn the axle today we're going to learn about sparkline sparkline is some kind of graph but
03:41instead of making like instant graph square it put the graph in a cell so for example let's see here
03:51okay so I've insert and this is the sparkline you can make like everything's 100 will do but I don't
03:59know I'm pretty tired to wait for it
04:02right okay I try to use this thing obviously text to speed to convert my persona has gained new power
04:15and voice to my persona has gained new power my models what a coincidence I hope it doesn't rain
04:26and the result is not still not really good because you know the model is usually like 600 epochs or
04:34more
04:35because your data is not that long either but it is what it is I guess it's okay it's it's
04:43starting to look like
04:52yeah it's it's more like it's okay now than my voice or other voices but it's still a way like
05:00100 epochs to go maybe
05:01anyway guys I'm not gonna use this video to train 100 more epochs
05:11but you know it is what it is thank you for watching don't forget to like share and comment
05:15and see you in the next episode have a nice day
Comments

Recommended