The evolution of the idea started out with a couple of wireframes that defined the core functionality at a high level coupled with two mockup screens. Through a series of discussion sessions and tech consultancy, we’ve proposed a set of solutions to implement for the MVP of the project. The proactive approach in constructing a plan for development has not only won us the client’s trust but also served as a good knowledge base for the early stages of development. 

Merixstudio’s team adopted the tested ScrumBan workflow and set up a full-stack development team consisting of an Agile project manager, experienced frontend and backend developer, and Quality Assurance Specialist. To efficiently tackle the project’s challenges, we extended the team with another 2 frontend developers, backend developer, Quality Assurance Specialist and DevOps Engineer. In accordance with the best communication practices, they stayed in constant contact with the client (leveraging different platforms such as Slack, Google Meet), exchanging ideas, confirming assumptions and consulting their work with the project’s stakeholders. 

The first big challenge for development (and one that was the initial focus for the team) was video generation using the user’s input. The file generation had to be smooth in terms of performance, as well as accurate in aligning audio files, video footage, background music and subtitle layers in one output file. To guarantee the scalability of the application, we have prepared the appropriate infrastructure and AWS environment as well as performed successful tests for generating 100 videos at one time.

Another functionality that was considered core priority was introducing Amazon Polly text-to-speech technology to allow for the creation of audio files from a user’s text script. The audio output files were then to be paired up with stock video footage selected by the application’s smart algorithms by extracting keywords from the script. 

Once the above have been taken care of, it was predominantly our Frontend Engineers that were tasked then with creating a bespoke editor for the videos users would create. Once all input has been generated, the application allows to manipulate it on asset-specific timelines and decide the succession of assets that would be used in the final cut. 

Lastly, our Python engineers carried the weight of developing an engine that would tie together all of the assets and their respective placement in time and generate one output video file to be then shared via social media channels. It was a big challenge to ensure that this is done smoothly both in terms of performance and alignment of audio & video files.