Making deepfake videos used to be hard. Now all you need is a smartphone. Avatarify, a startup that allows people to make deepfake videos directly on their phone rather than in the cloud, is soaring up the app charts after being used by celebrities such as Victoria Beckham.
However, the problem with many deepfake videos is that there is no digital watermark to determine that the video has been tampered with. So Avatarify says it will soon launch a digital watermark to prevent this from happening.
Run out of Moscow but with a U.S. HQ, Avatarify launched in July 2020 and since then has been downloaded millions of times. The founders say that 140 million deepfake videos were created with Avatarify this year alone. There are now 125 million views of videos with the hashtag #avatarify on TikTok. While its competitors include the well-funded Reface, Snapchat, Wombo.ai, Mug Life and Xpression, Avatarify has yet to raise any money beyond an angel round.
Despite taking only $120,000 in angel funding, the company has yet to accept any venture capital and says it has bootstrapped its way from zero to almost 10 million downloads and claims to have a $10 million annual run rate with a team of less than 10 people.
It’s not hard to see why. Avatarify has a freemium subscription model. They offer a 7-day free trial and a 12-month subscription for $34.99 or a weekly plan for $2.49. Without a subscription, they offer the core features of the app for free, but videos then carry a visible watermark.
The founders also say the app protects privacy, because the videos are processed directly on the phone, rather than in the cloud where they could be hacked.
Avatarify processes user’s photos and turns them into short videos by animating faces, using machine learning algorithms and adding sounds. The user chooses a picture they want to animate, chooses the effects and music, and then taps to animate the picture. This short video can then be posted on Instagram or TikTok.
The Avatarify videos are taking off on TikTok because teens no longer need to learn a dance or be much more creative than finding a photo of a celebrity to animate to.
Avartify says you can’t use their app to impersonate someone, but there is of course no way to police this.
Co-founders Ali Aliev and Karim Iskakov wrote the app during the COVID-19 lockdown in April 2020. Ali spent two hours writing a program in Python to transfer his facial expressions to the other person’s face and use a filter in Zoom. The result was a real-time video, which could be streamed to Zoom. He joined a call with Elon Mask’s face and everyone on the call was shocked. The team posted the video, which then went viral.
They posted the code on Github and immediately saw the number of downloads grow. The repository was published on 6 April 2020, and as of 19 March 2021 had been downloaded 50,000 times.
Ali left his job at Samsung AI Centre and devoted himself to the app. After Avatarify’s iOS app was released on 28 June 2020, viral videos on TikTok, created with the app, led it to App Store’s top charts without paid acquisition. In February 2021, Avatarify was ranked first among Top Free Apps worldwide. Between February and March, the app 2021 generated more than $1 million in revenue (Source: AppMagic).
However, despite Avartify’s success, the ongoing problems with deepfake videos remain, such as using these apps to make nonconsensual porn, using the faces of innocent people.
At first glance, Quiq and Snaps might sound like similar startups — they both help businesses talk to their customers via text messaging and other messaging apps. But Snaps CEO Christian Brucculeri said “there’s almost no overlap in what we do” and that the companies are “almost complete complements.”
That’s why Quiq (based in Bozeman, Montana) is acquiring Snaps (based in New York). The entire Snaps team is joining Quiq, with Brucculeri becoming senior vice president of sales and customer success for the combined organization.
Quiq CEO Mike Myer echoed Bruccleri’s point, comparing the situation to dumping two pieces of a jigsaw puzzle on the floor and discovering “the two pieces fit perfectly.”
More specifically, he told me that Quiq has generally focused on customer service messaging, with a “do it yourself, toolset approach.” After all, the company was founded by two technical co-founders, and Myer joked, “We can’t understand why [a customer] can’t just call an API.” Snaps, meanwhile, has focused more on marketing conversations, and on a managed service approach where it handles all of the technical work for its customers.
In addition, Myer said that while Quiq has “really focused on the platform aspect from beginning” — building integrations with more than a dozen messaging channels including Apple Business Chat, Google’s Business Messages, Instagram, Facebook Messenger and WhatsApp — it doesn’t have “a deep natural language or conversational AI capability” the way Snaps does.
Myer said that demand for Quiq’s offering has been growing dramatically, with revenue up 300% year-over-year in the last six months of 2020. At the same time, he suggested that the divisions between marketing and customer service are beginning to dissolve, with service teams increasingly given sales goals, and “at younger, more commerce-focused organizations, they don’t have this differentiation between marketing and customer service” at all.
Apparently the two companies were already working together to create a combined offering for direct messaging on Instagram, which prompted broader discussions about how to bring the two products together. Moving forward, they will offer a combined platform for a variety of customers under the Quiq brand. (Quiq’s customers include Overstock.com, West Elm, Men’s Wearhouse and Brinks Home Security, while Snaps’ include Bryant, Live Nation, General Assembly, Clairol and Nioxin.) Brucculeri said this will give businesses one product to manage their conversations across “the full customer journey.”
“The key term you’re hearing is conversation,” Myer added. “It’s not about a ticket or a case or a question […] it’s an ongoing conversation.”
Snaps had raised $11.3 million in total funding from investors including Signal Peak Ventures. The financial terms of the acquisition were not disclosed.
One of the challenges that some would-be TikTok rivals have faced is that they often lack the same robust set of content creation tools, like filters, effects, and tools for repurposing others’ content — like TikTok’s Stitch and Duet, for example. It now appears that Snapchat is working to correct that latter problem, however, as it’s been spotted working on a TikTok Duets-like feature called “Remix,” designed for replying to Snaps. This feature will allow users to create new content using their friends’ Snaps — a “remix,” that is.
Initially, the feature will allow users to reply a friend’s story with a remixed Snap. To do so, you can record your own Snap alongside the original as it plays — much like a TikTok Duet.
The feature, which Snap confirms has launched into external testing, follows Instagram’s public test of a similarly named “Remix” feature focused on Reels content. (It had also tested a version for Stories as a first step.)
In Instagram’s case, the company explains that Remix lets anyone create an Instagram Reel where your video and theirs play side-by-side. This is, essentially, Instagram’s own version of TikTok Duets, a tool that’s often used to interact with other TikTok users’ content. In Duets, TikTok users can sing, dance, joke or act alongside another user’s video; cook someone else’s recipe; record reaction videos; boost videos from lesser-known creators; and more. It’s a core part of what makes TikTok feel like a social network, rather than just a platform for more passive video viewing.
Last fall, TikTok announced it was introducing several new layout options for Duets in addition to the left-right layout, including a new top-bottom layout, a special “react” layout, and a three-screen layout.
Some of those same Duet formats and others now appear to be under consideration by Snap, as well.
In its Remix feature, Snapchat users are presented with a screen where they can choose from a variety of options for combining Snaps — including the side-by-side and top-and-bottom formats, as well as others like where content is overlaid or where you could react to a Snap.
Image Credits: Photo of Snapchat’s Remix feature via @alex193a on Twitter
According to reverse engineer Alessandro Paluzzi, who first spotted the addition, Remix also offers a way for users to tag friends or other people they want to have permission to either remix or share their Snap via a new toggle switch.
It appears that users will be able to access the “Remix” feature from the same menu where you can today either report” a Snap or send it to others.
This menu, of course, is also available from within Snapchat’s new TikTok competitor, known as Spotlight, launched last year.
Though initially, Remix is being tested among friends, we understand that it’s expected to make its way to other parts of the Snapchat app in time. And likely, this would include Spotlight. Much like TikTok, Spotlight offers a video feed filled with short-form, entertaining videos that you can scroll through with up and down swipes, often set to popular music — thanks to Snap’s music industry deals. This would be a natural fit for Remixes, as it’s a common way for users to interact with each others’ content to create a dialog.
Image Credits: Photo of Snapchat’s Remix feature via @alex193a on Twitter (opens in a new window)
Snap confirmed with TechCrunch it’s beginning to test Remix on its app.
“I can confirm that externally we are testing the ability to reply to a friend’s story with a remixed Snap,” a spokesperson said. “It lets you build on your friend’s Snap while recording your own alongside the original as it plays for contextual conversations on Snapchat,” they noted.
The company didn’t offer an ETA for a broader rollout at this time.
As enterprises around the world pour more money into research and invention to stay competitive, the need for analyzing the worthiness of R&D expenses also grows.
One company serving that function is PatSnap. When co-founder Jeffrey Tiong was working in the medical devices industry more than a decade ago, he realized how critical intellectual property and patents were in the tech world.
In 2007, Tiong launched PatSnap in Singapore to build a global patent search database and overtime pushed the firm into adjacent realms. PatSnap’s more recent software, which Tiong dubs “innovation intelligence,” helps enterprises analyze their R&D strategies, keep track of competitors, and identify potential partners by crunching data around the likes of scientific papers, government R&D grants and startup funding news.
“What we found is that a lot of companies [treat] innovation as a department, as a function, as a KPI in an organization,” said Tiong. “Many companies are hiring people… who have to find out what kind of technology is out there and who is doing what. You cannot do everything by yourself nowadays. You need to partner.”
Investors are paying attention to the R&D boom. In PatSnap’s latest funding round, the company attracted SoftBank’s Vision Fund II and Tencent as lead investors. The Series E round totals $300 million, with participation from CITIC Industrial Fund, which is part of Chinese state-owned conglomerate CITIC Group; Sequoia China; Xiaomi founder Lei Jun’s Shunwei Capital; and Vertex Ventures.
Masayoshi Son spent less than half an hour on a call with Tiong before the billionaire founder of SoftBank hammered out a deal for PatSnap. In his early twenties, Son invented and patented a device that he sold for $1 million, so “he understands the importance of inventions, IP and innovation,” Tiong said.
Tiong declined to disclose PatSnap’s post-money valuation in an interview with TechCrunch but said the number has crossed $1 billion.
The United States is PatSnap’s largest market, although China is rapidly growing as a revenue stream amid the country’s patent filing spree. In 1999, the World Intellectual Property Organization (WIPO) received just 276 applications from China. By 2019, that number rose to 58,990, surpassing that of the United States.
But compared with their western counterparts, Chinese corporations are less inclined to pay big bucks for software, which makes it challenging for SaaS companies to monetize in the country. PatSnap operates under the brand Zhihuiya in China, with customers ranging from retail brands, research institutes, AI firms to pharmaceutical giants.
The sheer number of patents doesn’t translate conveniently into technological clout. The U.S. is still ahead of China in terms of R&D expenditure, Tiong observed. Furthermore, “the quality of the patents in China is not as strong and a lot of them are increment innovation instead of groundbreaking types of invention,” he added.
PatSnap says it now has more than 10,000 customers in over 50 countries, with a 700-person workforce spread across the U.S., Europe, Canada, Japan and China. Some of its notable customers include Tesla, General Electric, Siemens, Dyson, PalPal, Spotify and Megvii. With the fresh capital, the company plans to further develop products, acquire more domain expertise, expand global sales presence and invest in human capital.
WeChat continues to advance its shopping ambitions as the social networking app turns 10 years old. The Chinese messenger facilitated 1.6 trillion yuan (close to $250 billion) in annual transactions through its “mini programs,” third-party services that run on the super app that allow users to buy clothes, order food, hail taxis and more.
That is double the value of transactions on WeChat’s mini programs in 2019, the networking giant announced at its annual conference for business partners and ecosystem developers, which normally takes place in its home city of Guangzhou in southern China but was moved online this year due to the pandemic.
To compare, e-commerce upstart Pinduoduo, Alibaba’s archrival, saw total transactions of $214.7 billion in the third quarter.
WeChat introduced mini programs in early 2017 in a move some saw as a challenge to Apple’s App Store and has over time shaped the messenger into an online infrastructure that keeps people’s life running. It hasn’t recently disclosed how many third-party lite apps it houses, but by 2018 the number reached one million, half the size of the App Store at the time.
From Tencent’s strategic perspective, the growth in mini program-based transactions helps further the company’s goal to strengthen its fintech business, which counts digital payments as a major revenue driver.
A big proportion of WeChat’s mini programs are games, which the app said exceeded 500 million monthly users thanks to a boost in female and middle-aged users, as well as players residing in China’s Tier 3 cities, WeChat said.
The virtual conference also unveiled a set of other milestones from China’s biggest messaging app, which surpassed 1.2 billion monthly active users last year.
Among its monthly users, 500 million have tried the WeChat Search function. The Chinese internet is carved into several walled gardens controlled by titans like Tencent, Alibaba and ByteDance, which often block competitors from their services. When users search on WeChat, they are in effect retrieving information published on the messenger as well as Tencent’s allies like Sogou, Pinduoduo and Zhihu, rather than the open web.
WeChat said 240 million people have used its “payments score.” When the feature debuted back in 2019, there was speculation that it signaled WeChat’s entry into consumer credit finance and participation in the government’s social credit system. WeChat reiterated at this year’s event that the WeChat score does neither of that.
Like Ant’s Sesame Score, the rating system works more like a royalty program, “designed to build trust between merchants and users.” For instance, people who reach a certain score can waive deposits or delay payments when using merchant services on WeChat. The score, WeChat said, helped users save more than $30 billion in deposits a year.
WeChat’s enterprise version has surpassed 130 million active users. Its biggest rival, Dingtalk, operated by Alibaba, reached 155 million daily active users last March.
The one-day event concluded with the much-anticipated appearance of Allen Zhang, WeChat’s creator. Zhang went to great lengths to talk about WeChat’s nascent short-video feature, which is somewhat similar to Snap’s Stories. He didn’t disclose the number of users on short videos because “the PR team doesn’t allow” him to, but said that “if we set a goal for ourselves, we will have to achieve it.”
Zhang also announced the WeChat team is weighing up an input tool for users. It’d be a tiny project given Tencent’s colossal size, but the project reflects Zhang’s belief in “privacy protection,” despite public skepticism about how WeChat handles user data.
“If we analyze users’ chat history, we can bring great advertising revenue to the company. But we don’t do that, so WeChat cares a lot about user privacy,” asserted Zhang.
“But why do you still get ads [related to] what you have just said on WeChat? There are many other channels that process your information, not just WeChat. From there, our technical team said, ‘Why don’t we create an input tool ourselves?'”