Good morning everyone am using w okada and when I load the model and I want to use it real time but the output isn't given me the required result because I trained a model with less than 15 minutes of the target voice and I want to know if training with more minutes will improve the model and what's the best way to train a good model that gives the required output?
#RVC model training
dusky lion
covert juniper
Good morning everyone am using w okada and when I load the model and I want to u...
Making a good realtime model is hard since inferencing happens realtime, more resources used
More training data wont help it, you need consistency across the entire dataset (same microphone used, same volume levels, etc.) you can make a good model with 5 minutes of data with that
the more variation there is the less realistic it will sound on realtime, thats just how it is. Thats why egirl, asmr, mommy voices sound good on realtime cause the dataset consistently sounds the same
dusky lion
the more variation there is the less realistic it will sound on realtime, thats ...
Can I inbox you privately please?
covert juniper
Can I inbox you privately please?
No, ask in here
dusky lion
Okay so I have the target video from a podcast which I used avidemux to remove the parts which are not the targets own and then used ffmpeg to convert it to a wav audio of 48KHz I think it the voice note amounted to about 30+ minutes don't exactly know the figure and then I want to train using RVC web and I don't know if there is anything I need to do before improving the model quality and then if the RVC web is best for training and I will be using w okada to test that because when I trained earlier which small dataset and I tried to use the RVC gui it keeps telling me to use a particular index file instead of another one despite the fact am using the exact index file it was talking about
So am really frustrated about the whole thing