dexotify, use code however you like

#4
by jvision - opened

Didn't text this implementation exactly, and know that you all require the rights release for any contributions. I share any copyrights with Coqui if you use the code. Understand if you don't use it or chose to re-implement it another way. It seems to be a quite capable method, for the small amount of use I see on the voices space I see a few rejections each day that seem appropriate.

I forgot to add the part that evaluates the response

    if (
            results['toxicity'] > thresholds['toxicity'] or
            results['severe_toxicity'] > thresholds['severe_toxicity'] or
            results['obscene'] > thresholds['obscene'] or
            results['threat'] > thresholds['threat'] or
            results['insult'] > thresholds['insult'] or
            results['identity_attack'] > thresholds['identity_attack']
    ):
        return "Input flagged as potentially harmful", 400
Ready to merge
This branch is ready to get merged automatically.

Sign up or log in to comment