Nice, I made something similar[1] but as a general language learning tool that works with 36 languages.
The main idea to improve your pronunciation is that you hear a native speaker (from a podcast or YouTube) and repeat right after, via Whisper ASR you get a score to see how accurate your response was compared to the native speaker. The method is based on the "language shadowing" technique.[2]
[2] https://en.wikipedia.org/wiki/Speech_shadowing#Language_lear...