INTERESTED IN INVESTING?
Where we're going
We are Synervoz Communications Inc. We develop cutting edge audio & communications software including the Switchboard apps and developer platform. Get in touch to discuss partnership, white label, and other opportunities.
Who we are
Next generation walkie-talkie and voice chat.
Switchboard is Superpowered
PLATFORM & SERVICES
© 2018 Synervoz Communications Inc. All rights reserved.
We'll get in touch shortly
How does it differ from PTT solutions like Voxer, Zello, AT&T's Enhanced Push-To-Talk (EPTT)?
Switchboard differs in many ways. Some of them are:
Switchboard is hands-free. Think of it as "talk-to-talk" (voice commands, not push).
If you hear someone talk through the intercom and respond, you are immediately in duplex (i.e. you aren't 'messaging' one at a time back and forth).
Switchboard is set up with persistent VoIP connections. This allows for fast toggling between voice channels using a variety of side channel controls (gestures, buttons, voice commands, 3rd party hardware, etc.).
Switchboard is the only platform with an always-on mode. This allows users to keep a connection running, and voice transmits only when they speak (using our proprietary Voice Activity Detector).
How does it differ from gaming chat like Discord, Teamspeak, Ventrillo, etc.
From communication style and technology perspectives, gaming chat is as similar as it gets to the Switchboard paradigm. But there are still major differences.
Gaming chat services do not persist channels in the same way, particularly on mobile, making them less flexible and and more difficult to toggle between channels.
The UI / UX is also completely different. You are generally connected to a particular "server" organized around a particular game, team/guild, or theme. Furthermore, you are only connected to one at a time. Gaming chat user interfaces make a lot of sense for gaming, but not the use cases we're focused on.
Gaming chat services do not have the audio mixing and ducking capabilities of Switchboard, especially on mobile. We have patented software that enables new experiences with music and for smooth mixing of voice with other audio sources.
There are additional challenges to running intercom-style and always-on voice chat in mobile and real world use cases. Noisy environments, less reliable internet connectivity, no paddle controllers / keyboards, a need to integrate with different services (e.g. office applications), different hardware (e.g. smart speakers, cars, smart earbuds, etc.). This makes the technology stack, and particularly client audio processing capabilities, very different. There is also a greater need for voice control and other side-channel controls (i.e. buttons, gestures, etc.) in the real world.
Does Switchboard require a proprietary network or can it be integrated on other networks?
We built Switchboard to be flexible in terms of its transport network.
We can route Switchboard voice traffic over any network and are happy to do custom deployments on customers' own networks; online and off-grid. Currently we have in-app toggles that allow us to switch between two networks on the fly. One of these is webRTC based, and the other is a client-server topology for use cases that can scale to an unlimited number of users per channel. We also have a working prototype where devices are connected via a mesh network.
What is meant by always-on / persistent VoIP.
There are nuances depending on the platform / settings, but generally speaking we maintain connections on the back end to enable quick set-up and toggling between voice channels. Depending on the specifics, socket connections may remain open with intermittent network traffic to keep them alive, or in other cases we 'wake up' a client that has been inactive for a period of time. From a user experience perspective, this means much lower latency than traditional VoIP applications. It is more like muting and unmuting oneself in an ongoing call than the many seconds it takes to set up a new call.
This should not be confused with "Always-On" as applied to things like Alexa, Google Assistant, Siri, etc. Always-On used in the context of voice assistants often leads to questions around privacy. This is a separate matter.
Are you doing speech to text and other voice services in-house?
We have developed some of our voice service technology stack in house and we are using open source or proprietary libraries in other cases. Where proprietary services are used they are cost effective with respect to our business model.
You mentioned something about machine learning, where does that fit in?
Our voice activity detection and speaker identification models are built in-house.
We have collected voice data (opt-in) and annotated it using a web-annotation tool that we developed in-house. The precise methods are proprietary, but we have worked with methods including recurrent neural networks, gaussian mixture models, vector quantization, support vector machines, ivectors, and utilized a number audio features including MFCCs, energy, spectral flatness, Zero Crossing Rate, and more. Because our use case is unique, we also have a unique set of constraints / trade-offs that have allowed us to work on unique approaches to solving voice detection and speaker identification problems.
How many users do you have?
This is in flux, please get in touch.
Why do users like it / what problems does it solve?
This varies per use case.
What's this about AI / voice assistant integration?
We have two broad ways that we are integrating AI:
1) Adding agents at the Switchboard contact list level (e.g. a weather bot, a Switchboard bot, a general purpose bot, a bot connected to a business' back end/database).
2) We are integrating with voice assistants at the hardware level (phone, speaker, vehicle, headphone, etc.) to gain access to the Switchboard network via voice commands such as "Hey Google/Alexa/Siri/Watson/Cortana/Bixby, is James available?"
What languages does Switchboard support?
Our in-house voice tech is based on audio features, not language. The third party services we use support multiple languages. Currently our UI, marketing, webpages, etc. are English only, and that is our main constraint to supporting other languages.
What do you / can you integrate with?
We have a Slack integration, and are currently working with a partner on our music integration. We remain focused on new features that are complimentary with respect to existing services (i.e. collaboration tools that enable text, calls, and video). For this reason we intend to partner, and even embed, where possible. We are in discussions with a number of partners.
Is there any hardware involved / required?
Switchboard is a software only solution. We can integrate with hardware to enhance the experience. For example, we have used buttons to control who you're talking to (flic.io). We have also looked at SDKs from headset manufacturers, smart speakers, and other devices (e.g. Thalmic Labs).
What platforms do you support?
We currently have an iOS app in production, macOS and web apps in beta testing, Android in alpha. Note that native apps are required in each case due to the low level audio programming necessary.
We also have prototypes with Smart speakers.
SDKs in development.
Do voice commands require training?
Yes, partially. Currently the user needs to train a few key wake words including "Hey Switchboard", "Respond", and "Over and Out". But when trying to connect to their friends, we recognize First Name _ Last Name automatically. We will improve this further as we advance our integration with AI based voice assistants and our conversational interface.
You mentioned Line of Sight and Proximity based controls. What are they?
Because we have made it easy to control Switchboard via side channels (i.e. our API structure makes it easy), we can use any number of things to modulate volumes, turn conversations on and off, etc.
Line of Sight refers to an example where, say, Person 1 looks at Person 2 and starts speaking. Person 2 can immediately hear. It could be "on/off" type functionality, or we could make it "louder/quieter" in conjunction with 3D / positional audio. Proximity based control would be something like, the closer I walk to you, the louder you hear my voice. Positional audio exists in the gaming world. With the right sensors it can be leveraged in the real world as well (e.g. imagine a construction site where people have headphones with sensors on them, rather than walkie talkies).
What is the music integration about / what does it have to do with voice communication?
Here are a few reasons voice chat & music belong together.
People wear headphones at work to avoid the noise of their neighbors (especially those who are not part of their immediate team). This necessarily cuts them off from people they DO want to communicate with. People do not always wear headphones to cut themselves off from 'everyone', they are often avoiding distracting conversation that does not involve them (this varies depending on role / team composition).
When people are listening to music, whether through headphones or through loud speakers, they are often doing something where they are open to being dropped in on. This is true at the office (for the reason above), as well as when they are at home (e.g. cooking, cleaning, hanging out).
Think of other experiences like being in a car listening to music. Sometimes you talk, sometimes you don't. It's a shared music experience with intermittent conversation. Like going to a concert together. Music has always been a social activity. There's no reason headphones or distance need to change that. Until now, technological constraints have limited the extent of remote social music experiences (e.g. listen together experiences use text chat rather than voice or video). Switchboard changes that thanks to cutting edge audio and voice tech.
Music is a great presence indicator. "Jim is listening to Led Zeppelin" -- in conjunction with other indicators like "on iOS", or "at home", or "with James" -- gives you a sense of what the person is up to before you drop in on them. In many cases it acts as an invitation. It is important to know that 'now is a good time' to drop in -- something I don't get with traditional calling apps.
We let users control their own availability easily, as well as their music/media status. Our unique statuses help users convey information easily to their network of friends / coworkers. We are also proactive in helping users to update their status. Nobody wants to remember to mark themselves as 'away' 10 times a day.
A few more points to keep in mind re: music.
What about safety when having both ears covered?
This depends on the use case, and in some cases we would not recommend wearing headphones. But "transparency" is a feature we built in to allow users to control the volume of real world sounds around them (e.g. traffic, sirens, etc.). This can also be turned on/off contextually (presets/modes) or automatically in response to sound detection (e.g. turn on transparency when a siren is heard, or when someone says your name). In many cases safety is actually improved in comparison to current alternatives. For example, anywhere that people text, or use a push to talk service, there are safety issues due to distraction and/or not being hands-free (e.g. dropping a tool from several stories up to answer a call). Furthermore, there are new devices that can be used to keep both ears open (bone conduction headphones, or see BoseAR for a new alternative).
What about military / police / medical use cases?
Switchboard is conducive to these use cases.
Depending on the use case, there may be additional requirements including specialized hardware, HIPAA compliance, mesh networking capabilities, among other things.
We are happy to co-develop solutions with customers and partners.
What about battery life and data consumption on mobile?
Depending on your phone and how often you use Switchboard, you may find that constantly keeping it open drains the battery in 5-20 hours (newer phones last longer).
The major consumer on our iOS app is the WiFi chip, and less so the audio processing / wake words. 'Always on' does not mean always-transmitting. Our voice detector (low power) allows connections to stay open but for data to be transmitted only during voice detection. This minimizes data consumption and power consumption.
Typically battery is not a concern with our desktop/web apps. In demanding mobile use cases like field services and construction, Mophie cases (or similar packs) help to mitigate issues. We are also in the process of optimizing for power consumption and expect major improvements starting in January 2019.
How do you make money?
We are currently focused on embedding Switchboard technology in other apps (B2B). This comes with service and licensing fees.
The Switchboard app currently serves as a demo and testing app, though there is an opportunity to generate B2C revenue via a freemium model.