Training args about rvc_cli HOT 13 CLOSED

you mean using audio split?

from rvc_cli.

The output result and also when I heard the audio that it divided into chunks it was also very robotic even if the audio split is false still the output has very robotic voice

from rvc_cli.

with what f0 method?

from rvc_cli.

rmvpe

from rvc_cli.

can you check with other models? i didnt have any problems, or if you can also upload the results to somewhere

from rvc_cli.

https://drive.google.com/drive/folders/1ZB1oz5j5AGA4eREWdD-8VH37yJZQJIOr?usp=drive_link
this is through rmvpe model
the params i used for training and inference-
python main.py train "PREETI" v2 50 False True 200 48000 8 0 True False False

python main.py infer 0 0 0.66 20 rmvpe "C:\Users\Aditi\Desktop\Aditi_work\RVC\RVC_CLI\target\female_english.wav" "output\preeti_fe_150epochs.wav" "C:\Users\Aditi\Desktop\Aditi_work\RVC\RVC_CLI\logs\PREETI_150e.pth" "C:\Users\Aditi\Desktop\Aditi_work\RVC\RVC_CLI\logs\PREETI\trained_IVF1643_Flat_nprobe_1_v2.index" True True

from rvc_cli.

ok, i’ll check it

from rvc_cli.

your model is only 150 epochs, did you try with an another hq model?

from rvc_cli.

used rmvpe till 250 epochs when i used it via gui 150 epochs were enough for good results

from rvc_cli.

I am also getting robotic results. I replicated the inference in the webui and the results were much better. Do you know what are we missing?

from rvc_cli.

That error occurs because the training parameters are not set correctly - I will make a new way to set the parameters, in the meantime you can use it in webui form using Applio.

from rvc_cli.

I get the same issue not only during training but a similar issue during inference also

from rvc_cli.

Check the last update with the new args system

from rvc_cli.