Google has announced the developer preview of TensorFlow Lite, a solution for enabling on-device machine learning inference with a small binary size and low latency. TensorFlow Lite is designed for mobile and embedded devices and should be considered an evolution of TensorFlow Mobile.
If it's speech-to-text services you're after, Vocapia Research is in the game, offering a variety of speech technology services and working with their clients to adapt, fine tune or create products that are relevant for specific applications. One of Vocapia's products is the VoxSigma speech-to-text software suite that is suitable for many languages and a variety of audio data types. The VoxSigma API is available as a Web service via a REST API over HTTPS.