Improvement Portuguese model for fluent speech

Hello everyone!

While Vosk is having an incredible accuracy, when I speak slowly and clearly, in a more "robotic way" (near 90% - 100% of accuracy), when I increase the pace to a more fluent, human and normal speech, it is not as accurate, identifying words that has nothing to do with what was said. I'd say 30-60% of accuracy.

My question is: is it possible to somehow improve the accuracy for a more fluent speech?

Some important info:
I'm speaking on the microphone, using _vosk-model-pt-fb-v0.1.1-20220516_2113_ and _vosk-server webrtc_.

Thank you!