Spokestack Launches “Voice First Mobile” Model to Make Mobile Voice Apps Perform Like Smart Speaker Skills

Voice technology platform Spokestack has debuted a way to make voice-enabled mobile apps perform like voice apps on smart speakers. The new system is launching in tandem with the company’s on-device natural language understanding engine, designed to speed up the performance of those mobile apps while keeping user data secure from the cloud.

Mobile Voice

Spokestack’s new system can essentially import an app used by Alexa or Google Assistant on a smart speaker or smart display to a mobile platform. The conversational model on an Amazon Echo or Google Nest device becomes the same one on the mobile device. The mobile app includes all of the elements of a separate voice assistant, including its own wakeword, intent understanding, and speech-to-text capabilities. It’s all integrated into the Spokestack Studio app, allowing clients to continue to build and tweak their app over time.

“What it does is extract the [voice app] from a smart speaker. It emulates it on the mobile app,” Spokestack CEO Mike Tatum told Voicebot in an interview. “The big ‘aha’ moment since our relaunch in January is that mobile devices don’t speak in the same conversational model as smart speakers. We thought, what if smart speaker developers had an easier time importing to mobile. This could kickstart a whole new type of mobile app, the voice-first mobile app.”

The smart speaker voice app on mobile device ideas acts as a complement for the platform Spokestack released in January. The revised Spokestack platform enables mobile app developers to include custom voices to speak with users. Those personalized voice assistants can sound like a brand spokesperson or whoever else the developer chooses, using recorded audio to build a voice profile that the AI can use to say sentences that were never recorded. The custom voice adds an extra dimension to the customized mobile app, just as the new model makes it perform like a voice app on a smart speaker. Spokestack’s mobile device system started with iOS, partly because it’s the more difficult platform, Tatum said, but the option will be extended to Android and potentially other mobile operating systems soon.

Local NLU

The other big addition to Spokestack is the on-device NLU engine. Usually, voice apps require interactions with the cloud to understand and respond to what people are saying. If the AI can run without consulting with the cloud, it eliminates the delay needed to transmit and receive data, while making the whole interaction secure from possible server hacks and allowing it to run even when the device doesn’t have a connection to the internet. Tatum compared the NLU model provided by Spokestack to the one Google incorporated into its Pixel smartphones as part of the Android 11 update. Google shifted all of Google Assistant’s navigation to the device, similarly accelerating interactions with users. Developers using Spokestack’s platform can garner the same benefits, he explained.

“All of the conversations happen on the device without going to the cloud,” Tatum said. “The developer can embed the Tensorflow-trained model on the app. Google has it, and some transcription models [stay on-device]. But we’re the first to allow anyone to build their own and put it on the app. It’s so much faster when then NLU model is on-board.”

To help promote the new features, Spokestack is running the Export to Independence contest for developers. The competition ends on July 15 and is aimed at younger startups and even individual voice app developers. Whichever contestant creates the best mobile port of a smart speaker skill, judged by a panel of voice tech experts, will win $3,000, with four runner-ups each nabbing $500 prize. Spokestack is also donating $5,000 to the Southern Poverty Law Center in the name of all of the contestants as part of the competition.


How Spokestack Gives Mobile App Their Own Branded Voice Assistant

Cerence Studio Gives Carmakers Advanced Voice Assistant Design Tools

Picovoice Debuts Platform for Building Voice Assistants Without the Cloud