Rivian debuts a pull-out kitchen for its electric pickup trick

Sometimes you need scrambled eggs. And with that thought, toady at the Overland Expo in Flagstaff, AZ, Rivian announced a major accessory for its electric pickup: A camp kitchen. The unit slides out from the Rivian R1T’s so-called gear tunnel that lives between the bed and cab. The kitchen includes storage and a stove that’s powered by the R1T’s 180kWh battery pack.

This kitchen unit is the first significant concept Rivian has unveiled for the pickup’s unusual gear tunnel. This space provides another locked storage compartment for the pickup — but why have it all, many asked when it was revealed? And now, with this kitchen unit, Rivian is responding to the questions. It seems Rivian wants to make its vehicles the center of an ecosystem of add-ons. The company already revealed racks, vehicle-mounted tents and even a flashlight that hides in the side of the driver’s door. Expect more camping and outdoor gear as Rivian cements its brand image around adventurers.

Rivian is positioning its products for a particular lifestyle. Think Patagonia-wearing, Range Rover-driving, outdoorsy types or at least those who aspire to have that image. It’s a smart play, and so far, Rivian has stayed true to this image. All of its advertisements, social media posts, and appearances make it clear that Rivian is carefully aligning its brand image.

Trucks and SUVs are generally marketed to workman and families. TV commercials feature dusty men hauling bails of hay and women unloading groceries and closing the rear tailgate with her foot. But not Rivian.

So far Rivian has shown its products in the backwoods, running trails and sitting next to campfires. The people in the commercials are on an adventure, wearing coats by The North Face and sleeping in REI tents. With the kitchen from today’s announcements, they can pull a kitchen out of their pickup and make some coffee.

Rivian tells TechCrunch this is just a concept, but the company intends to bring this unit to production. There are likely to be other units for the gear tunnel. I, for one, would love to have a slide-out dog washing and drying station because there’s nothing worse than putting a muddy dog in a truck.


Source: Tech Crunch

Magic Leap buys Belgian startup building hologram teleconferencing software

Princess Leia’s hologram message to Obi-Wan is getting closer to reality, at least augmented reality.

Magic Leap announced yesterday that they’ve agreed to acquire Belgian startup Mimesys. The team has been working on bringing Star Wars-esque volumetric video calls to the Magic Leap platform, and it seems that the Florida AR startup liked what they were developing. We didn’t get any details on deal terms.

The team is joining Magic Leap but will continue to service their enterprise clients including BNP Paribas and Orange, according to their website. The startup first showed off their video conferencing tech at CES this year, which allows someone in a Magic Leap One headset to visualize a 3D representation of a person during a “video” call.

Volumetric video can be fairly fickle, the solution Mimesys has been going after relied on Intel’s RealSense depth cameras to collect and stitch footage on PCs locally then stream it to a user’s headset. As is the case with almost all volumetric video footage, Mimesys’s early work appeared to suffer from some noisiness, though this acquisition makes one wonders whether Magic Leap will end up creating their own external depth camera hardware to appeal to enterprise customers.

It seems that almost every new platform promises to revolutionize video calls and yet hangups continue because its one of the more high-bandwidth activities we regularly do. This is only going to be more difficult in AR since you’re sending live 3D footage of people through the internet tubes. Like much of what Magic Leap is promising, the feasibility of pulling this off likely relies in no small part to 5G tech proliferating.


Source: Tech Crunch

“Old Town Road” finally gets the video treatment

From viral TikTok sensation, to music’s number one smash hit, “Old Town Road” has taken the top spot in music’s cultural firmament relying on an incredible Nine Inch Nails sample, some old fashioned controversy (it was barred from country music charts), and a remix with Billy Ray Cyrus.

So it’s only natural that the song that was created virtually and rose to number one through online ubiquity would work its way backward through the traditional music industry’s playbook to finally, after weeks atop the Billboard charts, get its video treatment.

And what a video treatment it is. This video has everything. Horse chases, car races, gunfights, time travel, and bingo.

Joining Lil Nas X is a cornucopia of famous faces from music and film including Billy Ray Cyrus (of course), Chris Rock, Diplo, Vince Staples, and Rico Nasty.

It’s been a long week, so take a break with a trip on the “Old Town Road”. It’s fun.


Source: Tech Crunch

Blue Origin and SpaceX get million-dollar NASA nod to test moon lander tech

Eleven aerospace companies will share more than $45 million in funds from NASA to design and test prototypes for the Artemis moon missions, the agency has announced. Among the established names like Northrop Grumman and Sierra Nevada are relative newcomers SpaceX and Blue Origin, looking to make a place for themselves on the agency’s biggest push in decades.

The funds are to enable what NASA calls undefinitized contract actions, in which partners get to work before negotiations on the rest of the contract have concluded. It basically shows that time is of the essence and that NASA is willing to pay up front to someone they may not even contract with later, just to get a jump start on the work that needs doing.

And what’s the work? They’ll be cooking up designs and prototypes for the Human Landing System, which as you might guess will take astronauts (and cosmonauts, and perhaps taikonauts) from a high lunar orbit to a low one, then to the surface, then back again. The three elements are called transfer, descent, and ascent respectively — and there’s a refueling one as well.

Each company will have a specific set of mechanisms or designs it will be expected to produce, but none is expected to put together the whole shebang.

“We’re keen to collect early industry feedback about our human landing system requirements, and the undefinitized contract action will help us do that,” said NASA’s Greg Chavers in a press release. “This new approach doesn’t prescribe a specific design or number of elements for the human landing system. NASA needs the system to get our astronauts on the surface and return them home safely, and we’re leaving a lot of the specifics to our commercial partners.”

In other words, this is still an information-gathering phase for NASA, though the contractors must consider it as potentially the first step in producing a major system for Artemis, so they can’t do anything by halves.

None of the companies was assigned design work on the ascender, which suggests the plans for that part aren’t as far advanced as the rest. SpaceX will be producing a study on the descent element, while Blue Origin is taking on studies for the descent and transfer elements, plus a prototype for the latter.

We probably won’t see these prototypes or studies any time soon — if they’re not chosen for production they may remain trade secrets for future bids or on the off chance NASA changes its mind. Even if they are chosen, they may have to go through several more iterations before they can be shown publicly.

Here’s the full list of companies and their responsibilities under the new funding:

  • Aerojet Rocketdyne – Canoga Park, California
    • One transfer vehicle study
  • Blue Origin – Kent, Washington
    • One descent element study, one transfer vehicle study, and one transfer vehicle prototype
  • Boeing – Houston
    • One descent element study, two descent element prototypes, one transfer vehicle study, one transfer vehicle prototype, one refueling element study, and one refueling element prototype
  • Dynetics – Huntsville, Alabama
    • One descent element study and five descent element prototypes
  • Lockheed Martin – Littleton, Colorado
    • One descent element study, four descent element prototypes, one transfer vehicle study, and one refueling element study
  • Masten Space Systems – Mojave, California
    • One descent element prototype
  • Northrop Grumman Innovation Systems – Dulles, Virginia
    • One descent element study, four descent element prototypes, one refueling element study, and one refueling element prototype
  • OrbitBeyond – Edison, New Jersey
    • Two refueling element prototypes
  • Sierra Nevada Corporation, Louisville, Colorado, and Madison, Wisconsin
    • One descent element study, one descent element prototype, one transfer vehicle study, one transfer vehicle prototype, and one refueling element study
  • SpaceX – Hawthorne, California
    • One descent element study
  • SSL – Palo Alto, California
    • One refueling element study and one refueling element prototype


Source: Tech Crunch

Pinterest delivers first earnings report as a public company

Pinterest (NYSE: PINS) shared impressive first-quarter financials on Thursday after the closing bell in what was its first earnings report as a public company.

The digital pinboard went public in April, rising 25 percent during its first day trading on the New York Stock Exchange. Pinterest’s public market performance has continued to stay in the green, closing up about 8 percent Thursday at nearly $31 per share for a market cap of $16.7 billion.

The company, led by co-founder and chief executive officer Ben Silbermann, posted revenues of $202 million on losses of $41.4 million for the three months ending March 31, 2019. This surpassed Wall Street’s revenue estimates of about $200 million in Q1 revenue and represents significant growth from last year’s Q1 revenues of $131 million. Losses, however, came in slightly higher than the expected adjusted loss of 11 cents per share at 32 cents per share.

“The IPO was a significant milestone, but our focus at Pinterest hasn’t changed,” Silbermann said in a statement. “We want to help people discover inspiring ideas for every aspect of their lives, from fashion and home decor to travel and fitness. Our success can be seen in our Q1 results, and we’re excited to continue to grow our reach and impact in the years to come.”

Pinterest in April sold 75 million Class A shares in an IPO that raised $1.4 billion. The IPO gave the company a fully diluted market cap of $12.6 billion, a figure slightly larger than its Series H valuation of $12.3 billion. This was amid concerns the company would see a slighter smaller valuation upon its IPO and gain the unseemly title of “undercorn.”

Pinterest previously disclosed revenues of $755.9 million in the year ending December 31, 2018, up from $472.8 million in 2017. Losses, meanwhile, shrank to $62.9 million last year from $130 million in 2017. For the full year 2019, Pinterest, expected to reach profitability by 2021, predicts full-year revenues of between $1.05 billion and $1.08 billion, up from $755.9 million in 2018.

Pinterest post-IPO performance and earnings report comes in stark contrast to both Lyft and Uber’s treatment on their respective stock exchanges. Lyft, for its part, has fallen since its IPO despite an initial pop of 21 percent. In its first-ever earnings report as a public company, released last week, posted first-quarter revenues of $776 million on losses of $1.14 billion, including $894 million of stock-based compensation and related payroll tax expenses. The company’s revenues surpassed Wall Street estimates of $740 million while losses came in much higher as a result of IPO-related expenses.

Uber suffered through a catastrophic IPO last week only to continue falling in the days since. The ride-hailing giant was previously valued at $72 billion by venture capitalists on the private market. It priced its stock at $45 a share for an $82.4 billion valuation last week. The company closed Thursday trading at about $43 per share for a market cap of $72.5 billion.

Pinterest’s disruptive digital advertising business appears to be more attractive to Wall Street than ride-hailing however. In addition to delivering an attractive earnings report, Pinterest displayed user growth. The company now counts 291 million monthly active users, a 22 percent increase from Q1 2018. Pinterest continues to gain global users, growing an impressive 29 percent in the last year. The U.S., however, remains the company’s core market where average revenue per user grew 41 percent to $2.25.

Pinterest was undeterred by skeptics, who predicted its nice-guy image and history of slower growth would make for a poor performing public company. Today, it’s market cap has surpassed Lyft, which was worth billions more before the two companies transitioned into the public markets.

How long Pinterest can stay in the green remains to be seen.


Source: Tech Crunch

Fiverr files to go public, reports revenue of $75.5M and a net loss of $36.1M for 2018

Freelance marketplace Fiverr has filed to go public on the New York Stock Exchange.

The company, which is headquartered in Tel Aviv, is losing money — its net losses grew from $19.3 million in 2017 to $36.1 million in 2018. At the same time, revenue grew by nearly 45 percent, from $52.1 million to $75.5 million.

“Our mission is to change how the world works together,” Fiverr says in the filing. “We started with the simple idea that people should be able to buy and sell digital services in the same fashion as physical goods on an e-commerce platform. On that basis, we set out to design a digital marketplace that is built with a comprehensive SKU-like services catalog and an efficient search, find and order process that mirrors a typical e-commerce transaction.”

Fiverr was founded in 2010 and, thanks in part to controversial marketing, is seen as a key player in the gig economy. It says it has facilitated more than 50 million transactions between 5.5 million buyers and 830,000 freelancers (who sell services like logo design, video creation and editing, website development and blog writing).

The company says its advantages include the breadth of the marketplace and a network effect where the number and success of buyers and freelancers on the site draws more buyers and freelancers. It also says its marketplace can be easily scaled up as it adds more freelancers from around the world.

As for risk factors, the filing points to the need to continue growing the community, the possibility that the overall freelance market may not grow as quickly as the company expects and he aforementioned history of losses.

Fiverr previously raised $111 million in venture funding, according to Crunchbase, from Bessemer Venture Partners, Accel, Square Peg Capital, Qumra Capital and others. It’s also made some acquisitions in recent years, including content marketing marketplace ClearVoice and And Co, which made software for freelancers.


Source: Tech Crunch

ObjectiveEd is building a better digital curriculum for vision-impaired kids

Children with vision impairments struggle to get a solid K-12 education for a lot of reasons — so the more tools their teachers have to impart basic skills and concepts, the better. ObjectiveEd is a startup that aims to empower teachers and kids with a suite of learning games accessible to all vision levels, along with tools to track and promote progress.

Some of the reasons why vision-impaired kids don’t get the education they deserve are obvious, for example that reading and writing are slower and more difficult for them than for sighted kids. But other reasons are less obvious, for example that teachers have limited time and resources to dedicate to these special needs students when their overcrowded classrooms are already demanding more than they can provide.

Technology isn’t the solution, but it has to be part of the solution, because technology is so empowering and kids take to it naturally. There’s no reason a blind 8-year-old can’t also be a digital native like her peers, and that presents an opportunity for teachers and parents both.

This opportunity is being pursued by Marty Schultz, who has spent the last few years as head of a company that makes games targeted at the visually-impaired audience, and in the process saw the potential for adapting that work for more directly educational purposes.

“Children don’t like studying and don’t like doing their homework,” he told me. “They just want to play video games.”

It’s hard to argue with that. True of many adults too for that matter. But as Schultz points out, this is something educators have realized in recent years and turned to everyone’s benefit.

“Almost all regular education teachers use educational digital games in their classrooms and about 20 percent use it every day,” he explained. “Most teachers report an increase in student engagement when using educational video games. Gamification works because students own their learning. They have the freedom to fail, and try again, until they succeed. By doing this, students discover intrinsic motivation and learn without realizing it.”

Having learned to type, point and click, do geometry and identify countries via games, I’m a product of this same process and many of you likely are as well. It’s a great way for kids to teach themselves. But how many of those games would be playable by a kid with vision impairment or blindness? Practically none.

Held back

It turns out that these kids, like others with disabilities, are frequently left behind as the rising technology tide lifts everyone else’s boats. The fact is it’s difficult and time consuming to create accessible games that target things like Braille literacy and blind navigation of rooms and streets, so developers haven’t been able to do so profitably and teachers are left to themselves to figure out how to jury-rig existing resources or, more likely, fall back on tried and true methods like printed worksheets, in-person instruction, and spoken testing.

And since teacher time is limited and instructors trained in vision impaired learning are thin on the ground, these outdated methods are also difficult to cater to an individual student’s needs. For example a kid may be great at math but lack directionality skills. You need to draw up an “individual education plan” (IEP) explaining (among other things) this and what steps need to be taken to improve, then track those improvements. It’s time-consuming and hard! The idea behind ObjectiveEd is to create both games that teach these basic skills and a platform to track and document progress as well as adjust the lessons to the individual.

How this might work can be seen in a game like Barnyard, which like all of ObjectiveEd’s games has been designed to be playable by blind, low vision, or fully sighted kids. The game has the student finding an animal in a big pen, then dragging it in a specified direction. The easiest levels might be left and right, then move on to cardinal directions, then up to clock directions or even degrees.

“If the IEP objective is ‘Child will understand left versus right and succeed at performing this task 90 percent of the time,’ the teacher will first introduce these concepts and work with the child during their weekly session,” Schultz said. That’s the kind of hands-on instruction they already get. “The child plays Barnyard in school and at home, swiping left and right, winning points and getting encouragement, all week long. The dashboard shows how much time each child is playing, how often, and their level of success.”

That’s great for documentation for the mandated IEP paperwork, and difficulty can be changed on the fly as well:

“The teacher can set the game to get harder or faster automatically, or move onto the next level of complexity automatically (such as never repeating the prompt when the child hesitates). Or the teacher can maintain the child at the current level and advance the child when she thinks it’s appropriate.”

This isn’t meant to be a full-on K-12 education in a tablet app. But it helps close the gap between kids who can play Mavis Beacon or whatever on school computers and vision-impaired kids who can’t.

Practical measures

Importantly, the platform is not being developed without expert help — or, as is actually very important, without a business plan.

“We’ve developed relationships with several schools for the blind as well as leaders in the community to build educational games that tackle important skills,” Schultz said. “We work with both university researchers and experienced Teachers of Visually Impaired students, and Certified Orientation and Mobility specialists. We were surprised at how many different skills and curriculum subjects that teachers really need.”

Based on their suggestions, for instance, the company has built two games to teach iPhone gestures and the accessibility VoiceOver rotor. This may be a proprietary technology from Apple but it’s something these kids need to know how to use, just like they need to know how to run a Google search, use a mouse without being able to see the screen, and other common computing tasks. Why not learn it in a game like the other stuff?

Making technological advances is all well and good, but doing so while building a sustainable business is another thing many education startups have failed to address. Fortunately, public school systems actually have significant money set aside specifically for students with special needs, and products that improve education outcomes are actively sought and paid for. These state and federal funds can’t be siphoned off to use on the rest of the class so if there’s nothing to spend them on, they go unused.

ObjectiveEd has the benefit of being easily deployed without much specialty hardware or software. It runs on iPads, which are fairly common in schools and homes, and the dashboard is a simple web one. Although it may eventually interface with specialty hardware like Braille readers, it’s not necessary for many of the games and lessons, so that lowers the deployment bar as well.

The plan for now is to finalize and test the interface and build out the games library — ObjectiveEd isn’t quite ready to launch, but it’s important to build it with constant feedback from students, teachers, and experts. With luck in a year or two the visually-impaired youngsters at a school near you might have a fun new platform to learn and play with.

“ObjectiveEd exists to help teachers, parents and schools adapt to this new era of gamified learning for students with disabilities, starting with blind and visually impaired students,” Schultz said. “We firmly believe that well-designed software combined with ‘off-the-shelf’ technology makes all this possible. The low cost of technology has truly revolutionized the possibilities for improving education.”


Source: Tech Crunch

Ticketmaster put an end to screenshots with new digital ticket technology

Ticketmaster is turning to new technology to help fight ticket fraud. The ticketing giant today unveiled its next-generation digital tickets, “Safetix,” which are tied to the ticket holder’s mobile device through an encrypted barcode that automatically refreshes every few seconds. The tickets will also support NFC technology, allowing fans to enter venues through a “tap and go” experience.

The company says ticket holders will later this year be able to add their contactless ticket to Apple Wallet, so they can enter a venue with their iPhone or Apple Watch. This will also involve the use of proximity-based technology which automatically selects the tickets when the phone is held near the ticket reader.

Apple and Ticketmaster already tested SafeTix this month during the fintech conference Transact, Ticketmaster says.

The combination of new technologies is meant to cut down on ticket fraud.

Today, unscrupulous resellers take screenshots or photocopies of tickets that they then sell multiple times over to unsuspecting victims. Because the barcodes now automatically refresh, a saved photo won’t work.

In practice, however, this may inconvenience some people who previously enjoyed the ease of screenshotting the ticket, then sending it to a friend — something that’s a lot faster than using the transfer feature on Ticketmaster’s website and in its app.

The change could also complicate things at venue check-in as users fumble with their phones to figure out how their new passes work — at least in the near-term.

For fans, the change means they’ll have to transfer tickets to friends, or anyone else they’re selling a ticket to, using the recipient’s phone number or email address. As a result, Ticketmaster gains visibility into the custody chain of each ticket, it notes. And that data can then be turned over to event owners, who will now have information about both the original ticket owner and the actual attendee, as well as anyone else who had access to the ticket.

This also means venues and event owners can target attendees with other offers and information about the event — like food, beverage or merchandise deals or venue-specific instructions. These are the fans they couldn’t have necessarily reached in the past, had the fan entered the venue using only a screenshot on their phone, for instance, or a paper ticket. The event or venue owner can even choose to follow up with the fan after the event wrapped, Ticketmaster says.

“Because a new ticket is issued every time there’s a transfer or sale, event owners have the ability to develop a unique relationship with each fan, leading to in-venue personalization and future communication while increasing their known fanbase,” explained Justin Burleigh, Chief Product Officer of Ticketmaster, North America, in a statement about the launch. “SafeTix will allow fans to arrive at a show or game with confidence that their tickets are always 100 percent authentic and will dramatically reduce the amount of ticket fraud event owners are dealing with on event day,” he added.

SafeTix aren’t immediately available everywhere, but are instead rolling out to specific venues and events to start. Initially, they’ll be used across NFL stadiums for the 2019 season and across a variety of touring artists’ acts.

They’ll later be available at Ticketmaster’s “Presence-enabled” venues — today that includes 300 venues across the U.S. where proximity-based technologies like NFC, RFID, and audio are used. (Of note: Ticketmaster partnered with TechCrunch Disrupt battlefield finalist Lisnr on the audio check-in functionality.)

SafeTix is only one way that Ticketmaster is leveraging technology at live events. The company is also now using facial recognition tools from Blink Identity, which it also invested in, at some venues. And Ticketmaster last fall acquired blockchain ticketing company Upgraded with an eye towards future enhancements of identity-based ticketing.

The transition to SafeTix shouldn’t be a major change for NFL game fans, however. Ticketmaster said that in 2018, 97 percent of fans entered venues using Ticketmaster technology during the 2018 season. It expects Ticketmaster Presence to be installed at over 350 venues in 2019.

 


Source: Tech Crunch

Rivals in gaming, Microsoft and Sony team up on cloud services

For the last two decades, Sony and Microsoft’s gaming divisions have been locked in all-out war against one another: on price, on hardware, on franchises, on exclusives… you name it. But it seems they’ve set their enmity aside temporarily that they might better prevent that filthy casual, Google, from joining the fray.

The official team-up, documented in a memorandum of understanding, was announced today, though details are few. But this is clear enough:

The two companies will explore joint development of future cloud solutions in Microsoft Azure to support their respective game and content-streaming services. In addition, the two companies will explore the use of current Microsoft Azure datacenter-based solutions for Sony’s game and content-streaming services.

Of course there is no doubt that Sony could have gone with a number of other cloud services for its gaming on demand services. It already runs one, Playstation Now, but the market is expected to expand over the next few years much like cord cutters have driven traditional TV and movie watchers to Netflix and other streaming services. Expansion would surely prove expensive and complicated.

The most salient challenger is likely Google and its new Stadia game straming service, which of course has a huge advantage in its global presence, brand recognition, and unique entry points: search and YouTube. The possibility of searching for a game and being able to play it literally five seconds later is an amazing one, and really only something Google can pull off right now.

That makes Google a threat. And Microsoft and Sony have enough threats already, what with the two of them making every exclusive and chip partnership count, the resurgence of Nintendo with the immensely popular Switch, and the complex new PC-and-mobile-focused gaming market making consoles look outdated. Apple Arcade exists, too, but I don’t know that anyone is worried about it, exactly.

Perhaps there was a call made on the special direct line each has to the other, where they just said “truce… until we reduce Google Stadia to rubble and salt the earth. Also Nvidia maybe.”

We don’t actually have to imagine, though. As Sony President and CEO Kenichiro Yoshida noted in the announcement: “For many years, Microsoft has been a key business partner for us, though of course the two companies have also been competing in some areas. I believe that our joint development of future cloud solutions will contribute greatly to the advancement of interactive content.”

Sony doesn’t lack technical chops, or the software necessary to pull off a streaming service — but it may simply make more sense to deploy via Microsoft’s Azure than bring its own distribution systems up to par. No doubt Microsoft is happy to welcome a customer as large as Sony to its stable, and any awkwardness from the two competing elsewhere is secondary to that. Google is a more existential competitor in many ways, so it makes sense that Microsoft would favor partnering with a partial rival against it.

Sony has long been in this boat itself. Its image sensors and camera technology can be found in phones and DSLRs that compete with its own products — but the revenue and feedback it has built up as a result have let it maintain its dominance.

Speaking of which, the two companies also plan to collaborate on imaging, combining Sony’s sensor tech with Microsoft’s AI work. This is bound to find its way to applications in robotics and autonomous vehicles, though competition is fierce there and neither company has a real branded presence. Perhaps they aim to change that… together.


Source: Tech Crunch

Google’s Translatotron converts one spoken language to another, no text involved

Every day we creep a little closer to Douglas Adams’ famous and prescient babel fish. A new research project from Google takes spoken sentences in one language and outputs spoken words in another — but unlike most translation techniques, it uses no intermediate text, working solely with the audio. This makes it quick, but more importantly lets it more easily reflect the cadence and tone of the speaker’s voice.

Translatotron, as the project is called, is the culmination of several years of related work, though it’s still very much an experiment. Google’s researchers, and others, have been looking into the possibility of direct speech-to-speech translation for years, but only recently have those efforts borne fruit worth harvesting.

Translating speech is usually done by breaking down the problem into smaller sequential ones: turning the source speech into text (speech-to-text, or STT), turning text in one language into text in another (machine translation), and then turning the resulting text back into speech (text-to-speech, or TTS). This works quite well, really, but it isn’t perfect; Each step has types of errors it is prone to, and these can compound one another.

Furthermore, it’s not really how multilingual people translate in their own heads, as testimony about their own thought processes suggests. How exactly it works is impossible to say with certainty, but few would say that they break down the text and visualize it changing to a new language, then read the new text. Human cognition is frequently a guide for how to advance machine learning algorithms.

Spectrograms of source and translated speech. The translation, let us admit, is not the best. But it sounds better!

To that end researchers began looking into converting spectrograms, detailed frequency breakdowns of audio, of speech in one language directly to spectrograms in another. This is a very different process from the three-step one, and has its own weaknesses, but it also has advantages.

One is that, while complex, it is essentially a single-step process rather than multi-step, which means, assuming you have enough processing power, Translatotron could work quicker. But more importantly for many, the process makes it easy to retain the character of the source voice, so the translation doesn’t come out robotically, but with the tone and cadence of the original sentence.

Naturally this has a huge impact on expression and someone who relies on translation or voice synthesis regularly will appreciate that not only what they say comes through, but how they say it. It’s hard to overstate how important this is for regular users of synthetic speech.

The accuracy of the translation, the researchers admit, is not as good as the traditional systems, which have had more time to hone their accuracy. But many of the resulting translations are (at least partially) quite good, and being able to include expression is too great an advantage to pass up. In the end, the team modestly describes their work as a starting point demonstrating the feasibility of the approach, though it’s easy to see that it is also a major step forward in an important domain.

The paper describing the new technique was published on Arxiv, and you can browse samples of speech, from source to traditional translation to Translatotron, at this page. Just be aware that these are not all selected for the quality of their translation, but serve more as examples of how the system retains expression while getting the gist of the meaning.


Source: Tech Crunch