This is a submission for the AssemblyAI Challenge : Sophisticated Speech-to-Text.
What I Built
Using the assemblyAI's test audio file, I transcribed a conversation between a weather podcast host and his expert guest discussing wildfires.
Using the AI's ability to detect individual speakers, I transcribed the audio using the utterance feature and labeled each speaker either 'host' or 'guest', respectively.
Rather than display all the text, I allow readers to paginate through sections of the conversation so they can read without scrolling on a standard-size computer screen.
Suppose a reader wishes to examine an individual question-and-answer in isolation. In that case, one need only click the speaker label associated with that section of the text, which will open a new web page where it can be viewed.
Demo
https://fierce-everglades-96194-97a0bfd171b5.herokuapp.com/
Journey
I used AssemblyAI's speech-to-text Model to transcribe an audio file into a user-friendly, readable format.
- used AssemblyAI to transcribe from an audio file and differentiate between speakers.
- rendered in a user-friendly, readable format.
- tested successfully for accessibility (tabs through nicely).
- creatively used pagination to break up the conversation into sections.
- let users focus on a single question-and-answer section.
Prompts
I accomplished the speech-to-text prompt with the provided audio file but did not use other prompts like streaming audio.
Team Member Submission
Just me - William Pope
Top comments (0)