wheat rock
What are some optimizations that I can do to reduce the cost of gemini 2.5 flash native audio? I have built a live interview platform and based on last month's analysis, I see that it takes around Rs 50-60 to run one interview (an interview lasts for 5-6 mins average), which seems to be very high.
I checked the input token usage and the maximum amount of input tokens that was used in one day was 3 million, and there were 12 interviews that day, which means on an average each interview used around 250K input tokens.
Any help on reducing the input token usage and the cost in general would be much appreciated 🙂