VoiceCue

Many of us have come across the tidy task of voice recording analysis, where you had to listen to the whole audio to identify the most essential parts.

Manual processing can be very time-inefficient. Just listening from end to end would often not be enough. You would have to double or even triple that time since you would have to pause and replay some parts of the audio.

I came up with an app that generates cue timecodes that lets you find all the important parts of your voice recordings like sentiments, entities, and tags with just a click.

Features

1. Voice recognition - based on the Deepgram
2. General stats - an overview of voice recording
3. Sentiment analysis - positive and negative word detection
4. Word cloud generation - most used word classification
5. Entity name recognition - categories such as person, place, etc
6. Activity tracking - find actions in past, present, or future
7. Interactive transcript - see progress or click to control it
8. Speaker detection - total number of speakers in recording
9. Cue word usage - short text samples for better context
10. Custom search - extended ability to query for cues
11. Waveform preview - see the dynamics of voice, identify silences
12. Audio controls - play, pause, fast forward, and backward
13. Drag and drop support - drop audio in the file select area
14. Upload MP3 files - the most commonly used audio format
15. Progress loaders - improved UX for loading transcripts
16. Fully responsive - works fine on mobile and tablets
17. Colorful UI - for easier interaction and word highlighting

Tech stack

The project is created using NextJS and Deepgram API.

The project is under the MIT licence and the code is available on GitHub.

Final notes

The full article on the making of the app can be read here.