Hi everyone, I’m using DeepGram Nova-2 (Italian transcriber) inside VAPI, but the results are not good — the transcription is inaccurate and not precise. However, when I test the same Nova-2 Italian model directly on DeepGram’s website, it works perfectly, word-for-word accurate. Why is there this difference in performance between VAPI and DeepGram’s site? Do I need to adjust some settings in VAPI for better accuracy? Also, if I were to create a custom DeepGram model/transcriber and host it on my own server, then integrate it with VAPI, would that solve the issue and give me the same accurate results @User @User