The photo art filter craze ignited by Prisma is extending to short video industry. Suddenly, a plethora of companies scrambled to launch their video style transfer tools. Facebook demoed its stylized filter feature this October, while Russian internet giant Mail.ru launched Aristo. Not to be left out, Prisma started to support real-time style transfer for livestreaming videos this week.
Philm, a video editing app that enables users to convert video clips into animated art works, is one of China’s answers to this trend. Philm, a combo from “photo” and “film”, adopts a neural network approach. Enabled by deep learning technologies, algorithms acquire the artistic style of a painting in terms of color and brush stroke techniques, and apply them in the creation of a new video.
What’s differentiates Philm from similar apps is its processing speed, according to Jiang Wenyi, CEO and co-founder of the company.
We all know that photo editing apps like Prisma take a couple of seconds to apply the filter because there’s a time gap for processing the images on the cloud and sending the treated pictures back to the users. “Through optimized algorithm, we’ve shortened the processing time with local rendering. Philm allows users to convert the video in real time even if there’s no network connection.” Jiang said.
The company’s CTO Zhang Yin told TechNode that the app could process videos at 30 frames per second (fps) on iPhone6s, generating more stable and fluid looking animations. “This is faster than Facebook’s soon-to-be-released Caffe2Go, which does style transfer at 20 fps on iPhone6s according to the press. Philm’s next update is expected to feature a higher frame rate of 40 fps.” he added.
The feature is currently only available to iOS users who get iPhone 5s or higher, owing to the spec requirements for running locally. The support for Android version will be launched in a few weeks.
In addition, the app features a lot of pinnable and resizable emoji stickers that move with your videos. After attaching the stickers to objects, they will move along with the targets, changing size and orientation to match the object they’re stuck to.
“The final results of these two features matched exceedingly well in style. Attaching a cartoon-style sticker to real world looking videos may seem somewhat obtrusive, but it’s harmonious when it appears in stylized video clips.” Jiang said.
Apart from obvious revenue sources from value-added services such as paid features and membership, the app plans to commercialize its service through cooperation with cooperates and IP brands. “Unlike basic filters, art filter technology has huge application potentials given that it can generate unlimited number of styles in a timely manner, For example, we could add Starbuck’s logo as a sticker or learn the style of Marvel Comics’ film trailers in a bid to promote their films.”
Also, filtered video adds a new expression to social networking by encouraging more people to create and share video content, according to Jiang. It’s applying a stylized mask to what you see for people can hide behind the filters to alleviate the social networking pressures, however the style you choose still reflects your tastes.
The app, which went live at the beginning of this November, is the work of a 30-employee team led by Jiang Wenyi, the company’s CEO and former co-founder of mobile app data analytics provider Umeng, and Zhang Yin, CTO and former associate professor at UT Austin.
The company has received $4.50 million USD of angel round from Innovative Works, Ping An Ventures, Trends Group, ZhenFund and Ceyuan Ventures. It is now looking for next round of financing, Jiang disclosed.