Cool, fast implementation of neural styling!
There are a bunch of image-modification apps out there, but it's interesting to see how a chatbot can fill all the same functionality. I see entertaining, content-focused bots as one of a few use cases working perfectly in a messenger interface. It makes me wonder what else chatbots can be taught to do, especially since developers can reach people on any platform.
And since text communication is so commonplace nowadays, I wonder whether Icon8 plans to scale to other messenger platforms, how their engagement compares between platforms, and how their engagement compares against conventional apps.
@bentossell thank you for the note. We do have an implementation for facebook messenger in beta. We are tweaking facebook specific UX and viral loop features to get it ready for launch. Will be happy to notify you when we are ready. Meanwhile please don't hesitate to provide your feedback on Telegram version)
@drawby Got it, we experiment with this workflow. We currently rely on that images in the thumbnail are at the same positions as buttons at the keyboard
Hey everyone — thanks for your upvotes, comments, and support. Happy to be on Product Hunt.
Here is some background on Icon8. It is a messenger-first neural styling for selfies. We’ve decided to focus on selfies as they are 30-50% of all pictures shared on social media. Our selfies and those of our friends are very dear to us all and are particularly important. So choosing this focus was easy.
Not so easy it was (and still is) to make great neural styles for faces. We specifically train our neural networks on faces to render facial features nicely. We also experiment a lot of various pre and post processing to make the styles just great.
The other interesting thing about us, is that we believe in simplicity, and that ‘less is better’. With this mind, we want to give each user only the very best filters which would work great for her. Not to make her choose from 20-30 styles, where only a handful would look nice. Instead, we give at one time only five to eight filters at one time and rotate them every 2-3 days leaving out those with poor engagement . To make this happen, our R&D process is quietly churning out hundreds of styles to make everyone’s filter collection useful.
All in all, Icon8 is built so that you get better and better filters the more you use it.
We are also big believers into messaging. We believe that messenger bots will win over apps in many use cases, including image enhancement. So far things go well, and we wonder how things will unfold.
What you do think? Will bots take over apps? What do you like\ dislike about out bot? Let us know.
Cheers,
Dima
@mikael_yang Hey, Mikael, sure, it's a part of the way product works. We rotate filters every 2-3 days. Each rotation we take away the filters users do not like or use and add new ones. In this way we gradually improve quality\engagement of filters for each user cluster. Eventually, each user would have her own unique set of filters.
@gabrelyanov I also love how filters work with landscapes.
At the same time we believe that the core is to provide the best experience with selfies first. If you get the best experience of making selfies artistic you will love the rest too.
Here are my favorite landscape shots:
@chikhalov thanx a lot! Love this question)
1) We already have beta on facebook m.me/icon8bot where we are tweaking UX and viral loop features.
2) Watsapp API is not awailable yet so not in our focus for now.
3) We do plan to launch icon8bot on other platforms such as: wechat, kik, viber.
Developing Icon8 on different platforms gives us an opportunity to get the best from each and share the knowledge across the platform
VC Puzzle