Product Design




Happy Scribe

  • JavaScript
  • React.js
  • React-native
Happy Scribe is a high-end dynamic SaaS platform for researchers, journalists, podcasters, and media production companies. It uses state-of-the-art machine-learning models to transcribe audio/video files uploaded by users automatically. Happy Scribe is used in a plethora of different industries, whether it be in education all the way through to the entertainment. To help them with the growth of their product, we were tasked with the development of a set of features.

Our cooperation with Happy Scribe came about thanks to our experience in other projects. As per their own feedback, they verified via our profile on Clutch that Merixstudio had an already validated background in this area of development. In the delivery of the project, all earlier assumptions and confidence in our capabilities were authenticated.

We were tasked with, among other features, the development of a simple and functional player for video or audio. Apart from simply playing the material, users can also control it by clicking on text to jump to the specific point in audio or video simultaneously.

When the media player was taken care of, the next thing we were responsible for making the transcribed text as clear for users as possible. That’s why we introduce speaker Identification, which allows users to set the identification to differentiate between speakers.

Another problem we needed to tackle was how to help users to easily navigate the text when playing the transcribed file. We achieved that by adding the function of highlighting the text (similar to karaoke), which is synchronized with the spoken word. In addition to that, we introduced comments, which also are based on highlighting particular parts of the text.

One of the most interesting features we implemented was also a confidence Heatmap, which defines the value in the audio/video file transcribed into text. Also, when users click the heatmap button, the words get colored depending on their confidence score - with that, words that have the biggest possibility to be incorrectly transcribed by the algorithm would be colored red.

Finally, all users work is private by default, but the application allows users to share a preview of the transcript. A user can export the final results to Word, TXT, PDF or SRT and VTT for the use as subtitles.

The technologies used to build the editor were JavaScript, React, SlateJS, and immutable.js libraries.

Visit site

We use cookies on this site to improve performance. By browsing this site you are agreeing to this. For more information see our Privacy policy.