When he first conceived of Momento, founder Genday Okrain set out to build an experience that would allow users to look back on memories since-passed. The result was a mobile app that could transform photos, live photos, and videos into shareable GIFs.
“The idea was to help people create sometimes complicated but high-quality things without them having to be professionals,” Genady said.
To add additional layers of complexity and quality without overwhelming his users, Genady realized Momento would need to offer additional creativity tools that were at once easy-to-use and powerful. Filters, effects, frames, sliders for contrast and GIF speed — these all seemed like good options.
But to add something that would truly differentiate Momento from the rest, Genady came up with the idea to add an augmented reality (AR) tool that would allow users to easily layer their gifs with digital objects.
These objects — things like hearts, music notes, and emojis — would open the door to more immersive, eye-catching creations that animate memories in unique ways that are perfect for sharing.
In order to fully implement this feature, Genady also needed a way to automatically identify and separate people in images from the background, so that the AR effects could combine with the scene more realistically.
To do this, Genady decided to implement a machine learning model that would segment out people and allow this dynamic interaction with AR effects. But the journey towards implementation turned out to be more difficult than expected.
ML solution: From in-house to off-the-shelf
Initially, Genady and the Momento team tried to develop their own solution in-house, but given the complexity and in-depth knowledge required to build and manage machine learning models ready for mobile, their initial efforts came up short. There were frequent bugs and crashes, and the team struggled to develop a model that they could confidently ship to their users.
So he and his team turned to external, managed solutions. Genady noted that ease-of-use, ML expertise, and the availability of a pre-built people segmentation model was what initially attracted him to Fritz.
“You need to have someone who knows what they’re doing,” he said, referencing Fritz. “If you can find someone who’s good at building these things and just use what they’ve built, that’s a good thing. And easy setup, good support. Those things are nice, too.”
Implementing Fritz as a solution also offered some additional benefits that promised to enhance the user experience.
“[Using a backend server/cloud] was something we wanted to avoid because of privacy concerns, price, and processing time,” Genady said, referencing the benefits of integrating the model on-device.
Ultimately, with Fritz managing the on-device ML solution, Genady and his team don’t have to worry about tweaking or updating the model. Instead, they can focus their time and energy on building features and experiences that will attract new users and bring those already using Momento back for more.
Making complex, powerful features accessible to users of all skill levels
When thinking about Momento’s users, Genady imagines small business owners or those trying to build emerging brands. These users, he believes, benefit from using professional-level creativity tools to craft engaging, high-quality content — without spending a lot of time and money.
“Think about someone who has no idea how to do these things,” Genady said. “It’s very hard to learn and do by yourself, but it’s easy to do automatically.”
Momento’s engaging user experience has also led to actionable feedback, which has helped in the development process. Genady noted that users love the AR features and have often reached out suggesting new effects for different seasons, holidays and events, and more.
Given this feedback, Genady and his team are considering the addition of new Fritz features like sky segmentation, which could enable users to bring sunsets of wisps of clouds to life.
This continued evolution of Momento goes to show that with on-device machine learning, the possibilities in building awe-inspiring creativity tools are truly limitless.
Comments 0 Responses