Skip to content

The local installation performs differently compared to the Hugging Face build #10

Description

@stunme

I installed BreezyVoice locally on a PC without a GPU, but the voice cloning results are noticeably different from the Hugging Face version. The local output has unnatural pronunciation, often fails to read the text properly, and sometimes blends the sample voice directly into the synthesized output.

The reference voice sample is 6–10 seconds long, and the input text for synthesis is around 300 characters.

Also, how can I set a random seed like the option available on the Hugging Face demo?

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions