Spaces:

GitMylo

/

bark-voice-cloning

like 98

Running

are you able to send me the audio which caused this error?

the audio is fine. the project itself has an error

i mean, i need the audio so i can know what's wrong with it, as any audio i've tried has worked, even if i do it right now. As you can see in the screenshot.

ok. here it is. this is a normal wav file
<audio controls src="{"error":"Only PNG, JPG, JPEG, GIF, MP4, MOV, QT, WEBM, MP3, MPGA, WAV files are supported"}">
um...

it seems that huggingface itsself is struggling with your audio file here as well? can you try ffmpeg -i audio.wav audio.wav (where audio.wav is your audio file name) to convert it and try that as the input?

Edit: Actually, i just looked at the error since i can apparently see the console. it seems that your audio gets loaded as a flat audio array, i'll make sure that it won't try to flatten a dimension which doesn't exist.

oh alright lol

I meant in a console, but scrap that, i just updated the app, try again now.

got it. it now successfully works with no errors and whatnot

great! i hope your audio works well too

update: i've tested out this model on google colab, and it's surprisingly well. it matched with the audio i was trying to replicate for. it's a 10/10 in my books :)

great!

Hello, thanks for your space, it worked for me. but how do i use it afterwards?

Edit: i was able to make it work i think? sharing the code for future reference: you use it in the generate_audio like this:

audio_array = generate_audio(text=text_prompt, text_temp = 0.8, waveform_temp=0.5, history_prompt="PATH/TO/GENERATEDFILE.npz")

Yep, that's the code to use it in bark in either raw python code, or a notebook