Voice cloning is not working (Error processing custom voice file: history prompt not found)

[agrogov]

I tried to reuse the sample custom voice from bark_voice_samples/female.wav and just got Error processing custom voice file: history prompt not found...
Is there any solution for this issue?

[jgravelle]

Yeah, sample voices are hard. You have to make your own history files, which is doable with the clone-able version of Bark: https://github.com/serp-ai/bark-with-voice-clone ...is a fork that supposedly helps make these *.npz files, but I blew an entire weekend with zero results. This thread looks promising: suno-ai/bark#382 If you figure it out, let me know...!

…

-jjg

On Wed, Oct 9, 2024 at 3:21 PM Alexey Rogov ***@***.***> wrote: I tried to reuse the sample custom voice from bark_voice_samples/female.wav and just got Error processing custom voice file: history prompt not found... Is there any solution for this issue? — Reply to this email directly, view it on GitHub <#6>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAZ6GXBTFYZHGCCJHX46I2DZ2WF23AVCNFSM6AAAAABPVJ673CVHI2DSMVQWIX3LMV43ASLTON2WKOZSGU3TMOBXGQ3TQMQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[agrogov]

I was able to successfully clone the example female voice with HuBERT, here is my modified notebook.
Just put generated female.npz in bark_voice_samples and that's it.

Changes in your code:

CUSTOM_MALE_VOICE_PATH = None
CUSTOM_FEMALE_VOICE_PATH = "bark_voice_samples/female.npz"

class GroqCasters:
    def __init__(self):
        try:
            self.groq = GroqProvider()
            self._setup_gpu()
            preload_models()
            self.custom_male_voice = CUSTOM_MALE_VOICE_PATH
            self.custom_female_voice = CUSTOM_FEMALE_VOICE_PATH

And remove _create_voice_prompt function, cos it's no longer needed.

[jgravelle]

Dude! This is huge!

Спасибо, товарищ...!

-jjg

[agrogov]

No problem!

And one more thing - you need to resample example voice to be 24KHz before, cos your initial file is 44KHz.