The Vision API is yet another pretrained ML API, like the video intelligence API. Check out how many different kinds of operations are possible with this: OCR, face detection, explicit content detection, and so on. You might be tempted to build an entire startup around this one, and you would not be the only one. However, it won't come cheap; performing millions of such operations will cost almost $80,000/month. Starting out is easy enough though, and if you are performing operations in the thousands rather than millions, the price will be just about $15/month: