Performance metrics, batching support, and French training support

[Alekksander66]

Hi,

Thanks for sharing your TTS project — great work.

I have a few technical questions to better evaluate it for production use:

What is the typical RTF (real-time factor) you observe in practice?

What is the average GPU and/or CPU usage during inference (and on which hardware, if possible)?

Is the model designed or well-suited for batch inference (multiple utterances per forward pass)?

Additionally, I’d like to know:

Is there any support or roadmap for training or fine-tuning the model in French?

If not, would the current architecture and training pipeline allow it with custom datasets?

Thanks in advance for your time and insights.

Best regards,