Almost a full year behind Spotify, US music streaming service Pandora has finally launched full voice functionality for smartphone users after completing successful beta trials earlier this year. It has a trick up its sleeve to trump its Swedish competitor, however, unveiling a hands-free feature which makes the Spotify feature, and indeed other in-app voice services, look redundant.
Like Spotify, Pandora’s voice functionality has been built in-house rather than relying on voice assistants like Alexa, Google Assistant, Cortona and even Bixby. But this as an interesting point because Spotify also has integrations with these proprietary assistants via smart speakers, while a key differentiator (a potentially negative one at that) is that when Pandora began trialing voice in January, it became the first pure play audio streaming outfit to offer in-app voice control and ignore integrating with the smart speaker voice assistant monopoly.
We can’t help but feel Pandora has missed a trick here, but it clearly sees some value in operating outside of smart speakers, as well as flaunting this hands-free feature which can be activated with a simple command, while Spotify’s in-app voice feature requires a long-press of the search icon.
As we noted at the time of Pandora Voice going live in beta mode, the once troubled US internet radio darling has decided the voice opportunity is too great and therefore tightly integrating voice into the app itself without Alexa is the way forward, becoming the second company to do so only after Amazon with its Music app. However, Pandora used its original announcement to take a dig at Alexa, claiming its new Voice Mode feature comes with more functionality than Alexa – including open-ended queries, interactive requests and directional requests. Pandora avoided adding any such jibes into this week’s full roll out announcement – maybe Alexa took notice?
So, what can we dig up about Pandora’s back office and natural language processing clout compared to Spotify’s? Six months seems like a long time from beta to full rollout but in that time a killer amount of data will have been collected and processed by Pandora.
Pandora’s Voice Mode is built on a voice and conversational AI platform from song recognition firm SoundCloud called Houndify, along with Pandora’s proprietary Speech-to-Meaning and Deep Meaning Understanding voice recognition and understanding technologies. These confusingly-named terms can be explained as we can create representations that encode aspects of what we would understand as meaning – such as basic sentiment, speaker-attributes and other fairly basic and well-defined things that are studied in the field of natural language processing. As for achieving “meaning” as we would know it, we have it on good merit from experts in the voice technology sector that the market is a million miles away from cracking the meaning nut.
Meaning can relate to picking up and processing notoriously tricky conversational nuances like sarcasm and irony, or it could pertain to no accent being too strong to process and understand, and perhaps even the seemingly impossible feat of mastering the art of every single English language dialect (there are approximately 160).
Little else has been revealed about the guts of Pandora Voice although the company has highlighted a focus on vague utterances which it believes will only become more and more vague as consumers adapt to interacting with technology via voice commands, akin to how they would with friends, according to an interview with Variety.
Pandora Voice supports all the commands you would expect, as well as more personalized requests such as “Hey Pandora, play my morning music”, which will initiate a playlist based on previous content played at a certain time of day. Maybe in the future, voice services will be capable of determining mood based on the tone of voice, although this hasn’t been mentioned as a desired feature for either Pandora or Spotify.
Another recent and arguably more ambitious voice frontier saw Pandora begin testing interactive voice advertising in April with project partner Instreamatic – essentially to encourage users to respond to ads. “A listener is delivered an ad with a verbal call-to-action to learn about new features of a new smartphone. The listener can respond to that ad simply by speaking aloud – either affirmatively to get more information or negatively to skip the ad,” is the example provided by Pandora. We can already hear the cries of expletive-riddled responses to skip unwanted ads.
The Instreamatic.ai platform allows media companies and advertisers to measure, manage and monetize content using natural language understanding designed to study, interpret and understand user intent, while continuously refining this through deep learning mechanisms. Pandora says the platform continuously learns and increases its vocabulary and predictive power and claims Instreamatic’s platform offers evidence that allowing consumers to have a dialogue with ads – yielding richer engagement and more powerful commitments to ad creative.
We wonder how much SiriusXM has had a say in Pandora’s voice technology initiatives since buying the company for $3.5 billion in October last year. Pandora has had a rough ride in recent years, but we commend the company’s R&D efforts to transform the way consumers interact with technology and brands. It hasn’t given any guidance to when its interactive voice advertising feature will go live, although if we apply the six-month rule as with Voice Mode, we could see it roll out around October.