interaction with boats cop shyam Vedanta
Earbuds have traditionally been simple audio devices. They play music, handle calls and, in some cases, offer features such as active noise cancellation and spatial audio. With the addition of AI, however, brands are now looking at earbuds as something that can do more than playback audio.
Boat is one of the brands exploring this shift with Crest AI. Built around Google Gemini and Gemini Live, the platform lets users talk to the AI through compatible earbuds and use voice for tasks that would usually require a phone. This includes summarising emails, managing calendar events, searching for information, translating conversations and finding music based on what they want to hear.
The company says it has designed Crest AI as a cloud-based experience rather than putting heavy AI processing directly on the earbuds. This is particularly important for Boat, given the price-sensitive segment in which many of its audio products compete.
We spoke to Shyam Vedantam, Chief Product Officer at Boat Lifestyle, to understand why the company built Crest AI, how it handles conversations in noisy environments, what happens to user conversations, and where the company sees AI fitting into its audio ecosystem.
Q: Why did Boat decide to build Crest AI, and what exactly is it designed to do?
Shyam Vedantam: The timing is right for building AI-enabled consumer experiences, and we believe voice is a natural form of user interface for devices such as personal audio. Crest is essentially our umbrella platform for these AI experiences.
What Crest AI does is transform audio devices from passive playback devices into active, context-aware AI companions, through our integration with Google Gemini and Gemini Live. We have launched a range of AI-powered smart audio devices that can listen, understand, converse and assist users.
One of the key differences from traditional voice assistants is that they are generally designed around rigid, single-turn commands. It is more of an answer-and-response system. Crest AI is designed to operate as a continuous, screen-free conversational interface directly through earbuds or headphones.
For example, we have a feature called G Assist, which is our productivity agent. It has tight integration with Google Cloud and the user’s Google account, allowing users to perform multi-step tasks related to email, calendar and other activities. You can summarise emails or create calendar events using your voice without taking your phone out of your pocket.
There are also use cases around entertainment, music, translation and device control.
Q: Earbuds are often used in noisy environments such as gyms, streets and public transport. How does Crest AI handle ambient noise, accents and speech input?
Shyam Vedantam: Crest AI is designed to work in noisy ambient environments, which is particularly relevant in India because indoor and outdoor environments can be inherently noisy.
The devices running Crest AI have powerful AI-based noise suppression capabilities. This reduces ambient noise and ensures that voice data is filtered and passed on to the AI models. We also use voice activity detection, which essentially detects what the user is saying, separates it from ambient noise and passes the relevant speech to the AI models.
All our voice agents today are built using Gemini Live. It provides real-time speech and streaming communication with low latency.
Also read: Our most intelligent machine: Dyson’s Tim Hare on V16 Piston Animal vacuum cleaner launch in India
Currently, we primarily support English language input, where the user speaks in English and receives feedback in English. We are also working on multilingual speech interaction, including Indian languages, which is currently under development.
We also have a feature called Crest Translate, which supports bidirectional translation across multiple Indian and international languages. However, command input is primarily in English at the moment.
Q: Are the conversations users have with Crest AI stored on the earbuds or elsewhere? Can users access their previous conversations?
Shyam Vedantam: As of now, we don’t store conversations on the devices. The entire conversation with the cloud agent is encrypted and sent to the backend. It is not stored on the device for data privacy and security reasons.
The experience is completely real-time.
Q: AI features can add hardware and processing costs. How has Boat managed to bring Crest AI to earbuds that are targeted at the mass market?
Shyam Vedantam: Making sure we could bring Crest AI to devices that are affordable for most Indian consumers was an important design criterion for us.
That is why Crest AI was designed to run as a cloud experience. We did not want to increase the cost of the devices by incorporating specialised models directly on the hardware. We kept the device-side agents as lightweight as possible and designed the service to run through our companion app and the Crest AI cloud.
At the same time, we wanted to make sure this architecture did not compromise the features users expect from their earbuds. Users are particular about playback time, sound quality and connectivity, among other things.
The AI experience was therefore designed so that it does not compromise these aspects, while still delivering voice experiences and keeping the overall cost of the device as low as possible.
Q: Will users have to pay separately for Crest AI?
Shyam Vedantam: There are currently non-Crest AI and Crest AI variants of the products. The Crest AI packs are available for Rs 399 for a one-year subscription.
For the initial launch, the subscription will be priced at Rs 399 for a year. We are not sure yet how the pricing of the subscription could evolve in the future.
Q: Beyond having a conversation with the AI, what can users actually do with Crest AI?
Shyam Vedantam: Crest AI has a suite of agents. There is Crest Ask, where you can brainstorm with the Crest AI agent. It is not limited to a simple query-and-command response. You can also ask for complex, multi-source information retrieval.
Then there is the translation agent, which handles bidirectional translation across multiple languages. We also have a device control agent, which can operate different aspects of the device. For example, users can ask it to increase or reduce the volume.
Another area is music. We have integration with YouTube Music, where users can create curated playlists and search for tracks based on mood, language or genre, depending on what they want to listen to.
Finally, there is a device troubleshooting mechanism. Users can ask questions about how their device works, such as how to turn on ANC, enable spatial mode or activate head tracking on supported devices. The agent can explain how to configure the device or enable a particular setting.
These are some of the agents we have for the initial launch.
Q: Boat has been selling audio products in India for years. What has the company learned about Indian users that has influenced the way Crest AI is being designed?
Shyam Vedantam: Over the last several years, we have developed a deep understanding of the different situations in which our products are used in the real world. This has given us insights into the typical usage patterns of Indian consumers.
Over time, we have internalised several aspects of how Indian consumers consume content, where they go to consume it, what kind of content they consume, how they shop online and offline, and what apps they use for things such as food delivery and payments.
There are multiple aspects that we have internalised through our research, not just around our products but beyond them as well.
Our thought process is to bring these unique Indian experiences into the overall Crest experience and make them an integral part of the platform, seamlessly blending them into Crest over time.
Q: Do you see Crest AI eventually expanding beyond earbuds and personal audio devices?
Shyam Vedantam: Right now, Crest AI is available on personal audio devices. Wherever voice as an interface mechanism can find a natural fit, we will definitely explore the possibility of integrating these kinds of voice-led experiences into other surfaces.
Keep reading Digit.in for similar interactions.