How should I implement sentiment analysis with no money?

I'm making a product and hosting on a very cheap server I can afford. One feature uses the sentiment of user input.

So far, I'm doing it with an old, bulky neural network running as a process on the same web server. It works okay, but I think that will scale very poorly.

So I'm considering other options. I can use a small, fast naive sentiment classifier (like TextBlob), but it has really bad accuracy. Alternatively, I can use free credits on the Google Cloud sentiment API until I get a customer and can afford to set my own fast, accurate analysis server up. But I don't know if I will have enough cloud credits to get me to that point, without getting squeezed.

So I'm not sure what to do. Do you have any ideas?