Inspiration
Our desire to learn and master a foreign language and our love for films.
What it does
It helps user learn a language by prompting questions for the user based on the captured audio.
Challenges we ran into
Challenges and Solutions:
One of the problems we encountered initially was that the transcription prematurely started before the disk could capture the full audio file. To overcome this, we implemented threading and made flags to make sure the recorder and the Whisper processor did not overlap.
API Latency: One problem we had was fine tuning the duration of the audio recording because although this provided the AI with more information, it would cause longer response times making the app feel slow and in some cases errors. To solve this we optimized the Google GenAI to limit the response tokens making the feedback feel snappier and tested around the durations of the recording repeatedly to discover the sweet spot.
Log in or sign up for Devpost to join the conversation.