Clutter Cleaner

• 

  Before cleaning

• 

  After AI has cleaned the image

Inspiration

As a work-from-home parent of an 8 month old, my background is often filled with clutter - baby toys, books, clothes, etc. I wanted to build a background generator that uses your existing background, lighting and furniture to give the most natural looking background that is clutter-free. The in-built backgrounds of the various meeting platforms look unprofessional due to how fake and choppy they are when used alongside the webcam.

This situation inspired the idea to create an tool that can take a picture of my background and then using the power of AI cleans up the clutter. Hence ClutterCleaner was born.

My situation isn't unique. People who work from home often need to attend online meetings, and really who wants to go through and clean their house every time they have a meeting. Most people end up using these fake backgrounds or blur their camera, neither of these look great. Enter ClutterCleaner, the tool that helps clean up the background in a simple way.

What it does

Clear the clutter from your real background so that your video meetings have the most natural background without any of the issues and limitations of a fake background/green-screen. The app will allow users to take a picture of their background with the webcam they will be using for video calls. This picture will then be cleaned of any clutter and can also be enhanced to match their business, the meeting, the pitch etc. Every time you are about to sit for a meeting, head to ClutterCleaner, take a picture and set that as your background. It's quick, simple and makes your meetings look professional. The reason you should take a picture every time, is to match the current time and lighting conditions for the most professional look.

How it was built

The App was built almost entirely with Bolt. Using the integrations with Supabase and Netlify. Google's Vertex AI Studio were used to identify the different AI Models and fine tune the prompt to get the best kind of outcome that stays true to the objective of ClutterCleaner. Other tools like Mailgun and Google Analytics are used for specific things like contact forms and the website analytics.

Challenges

The image generation AI would not provide a very good result. The recommended AI model would regenerate/reimagine the entire background rather than decluttering the original background, which was not my intention. After using bolt AI and my own research to find alternative, I was able to identify an alternative. However, the new AI was not working the way I hoped and I soon hit a road block. Thankfully I was able to overcome this challenge by manually debugging and solving the issue.

The current Image Generation AI being used has a low quality output, I would like to explore more options where the users can get an HD image or even have more refined control on the output. This is still an ongoing challenge I need to find a solution to.

Accomplishments

I was able to find a solution to the AI Generation issue and was able to have a UI and image generation that I'm happy with.

Overall I'm very happy and proud of what I was able to accomplish with Bolt. After facing a few roadblocks and some scary moments, Bolt has been the perfect companion in the journey.

Lessons learned

The original idea was specific towards me, but during the research and building phases I realised the bigger picture. There is much more to my struggle and many people in different situations can relate to it. This is a product that I have been thinking about building for some time, the hackathon was exactly the kind of push I needed to get things off the ground. This has been the biggest learning, don't over think it, jump in.

What's next for Clutter Cleaner

In the future, I would like to build ClutterCleaner to work as a plug-in with most meeting softwares so that the background generates automatically before every meeting. This way the background can exactly match their situation, the time and room lighting.

An extention like this can also allow the background to be updated/refreshed at a pre-determined time interval to account for lighting changes. This handy if the meeting is longer or at a time where the lighting changes frequently (e.g. dawn, dusk, noon).

Here are a few other points that I would like to address in the future:

• As mentioned above, the current image quality needs to be improved, I would also like to add functionality for users to have refined control on the output.
• Ability to manually select problem areas in the image or space and have the AI model work with these inputs as well.
• Explore further monetisation and subscription option which adds more functionality at a premium price. (Some of these features have been discussed in this section)
• Being able to schedule jobs and respond faster by giving priority to paying customer.
• Build a complete and functional dashboard to manage users etc as I currently only rely on the Supabase integration.
• Potentially add the ability for users to add some comments, which can be added to the base prompt in order to change the outcome of the cleaning.
• Referral system. Very similae to Bolt's own referal system, for every friend that joins you get X credits, but for every premium member you get Y additional credits.

Built With

• ai
• bolt
• react
• supabase
• tailwind