Open Interval

36h Hackathon

Overview

An AI-powered interval timer platform created during the Elevenlabs Hackathon, enabling creators to craft complex interval sequences with high-quality voice guidance.

Key Achievements

  • Achieved working prototype on short deadline
  • Integrated text-to-speech with cloud storage
  • Built intuitive interval sequence editor

Project Scope

  • Not intended to become a major project
  • Done for the experience of a hackathon, and for the learning under pressure opportunities

Technical Details

Tech Stack

Next.jsTypeScriptTailwind CSSRadix UIElevenlabs TTS APIUploadThingNeon Postgres DBDrizzle ORM

Design Tools

Not required

Key Features

  • Custom interval sequence creation
  • High-quality voice label generation
  • Audio label auto-generation
  • Public sharing of interval sequences
  • User authentication

Challenges & Solutions

Challenge:Batch processing audio requests, and getting the .mp3 files stored somewhere
Solution:UploadThing API enabled a backend workflow to submit requests to Elevenlabs and then immediately upload the mp3 files to Upload Thing
Challenge:Creating a decent editing experience
Solution:Enabling the user to dump their own JSON file into the editor, and essentially use an AI chat for changing the JSON file was the best solution for editing
Challenge:Creating an audio player that can stream multiple channels of "intervals" with audio files spaced out in time
Solution:This one was not entirely solved

Lessons Learned

  • Building the multiple audio streams / audio layers is hard, especially when the interval durations and the audio file durations do not match each other. To build a great product with this concept would require very good design around how the interval player works under the hood.
  • Text to speech AI credits are prohibitively expensive to allow just anyone creator level access to the platform. Eleven Labs still has the highest quality TTS technology, but it would be better utilised in a way that limits people from creating non-stop audio experiences like 10-30 minute guided meditations, for example.
  • Text to speech audio is at a point where it's very good, and there is a type of formatting language similar to mark down that helps to guide the AI for tonality, pausing, pacing etc.
Open Interval