conda activate hack
To run:
python main.py
then select your input audio (0 most of the time)
This uses 2 OpenAI API
Step 1: Capture Audio -> get transcript Using pyaudio
Captures the device's audio input as a wav file.
Step 2: transcript -> classification [0,1] value
The wav file gets send to openAI speech to text to get the transcript.
The transcript text is passed to openAI assitant instructed to provide a single confidence on if the text is an ad read.
Step 3: control the device
if ad read then decrease volume and birghtness. Bring it back once the ad is over.