Microsoft Translator: Speech Translation Made Easy
File it under the category “The shape of things to come.” Only, it’s here already.
Many folks have heard of the release of Microsoft’s Skype Translator feature. Previewed in December 2014 and launched in late 2015, it has already enabled the translation of millions of live conversations between individuals who could not otherwise communicate. While quirky to set up, a bit slow in responsiveness, and still far from perfect, it is an evolutionary step towards the sort of Universal Translator envisioned in Gene Roddenberry’s Star Trek, or the Babel Fish of Douglas Adams’ Hitchhiker’s Guide to the Galaxy.
The advantage of such a tool is that it provides immediacy to communications. The disadvantage is that it produces imprecise translations. It will fill a certain niche for communications, but do not be oversold on the concept: professional live translation will still be required when matters of etiquette and linguistic precision are paramount.
Starting in late March 2016 Microsoft “popped the hood” off the Skype Translator engine and began to offer the Translator Speech API as a standalone service offering. Initially offering eight languages, Microsoft plans to add more languages over time.
The API service is available through the Azure Marketplace. Pricing is free for up to 7,200 transactions per month (about 2 hours of conversation) which may be sufficient for initial development and testing, then rises to $120 per month for up to 36,000 transactions (10 hours), and $1,000 a month for 360,000 transactions (100 hours). At this highest tier, pricing works out to about $10 per hour for the service.
In a YouTube video just released by Microsoft Research, Chris Wendt and Kelly Altom walk developers through the Speech API, showing the application flow, how to establish a web socket, how to package and ship your audio and handle the returned information to and from the server, and in the end demonstrate a sample application.
The iOS Swift code for the sample “Tourist Information Center app” is over on GitHub, and the documentation for the Translator APIs is also online.
What are your thoughts on Microsoft’s automated translation APIs? We’d love to hear your feedback. Send us your thoughts at projects@e2f.com.