Technology

Reforge Audio Engine

Patent Pending. Built by Forge Hut LLC.

← Back to all research

Built for one purpose

Each soundscape in Ebbi has its own research page with the evidence behind it. Reforge is how we build them, and below we say which ones we tested and how.

Most focus and sleep apps ship pre-recorded audio loops downloaded from stock libraries. The same rain sound you hear in one app is the same file sold to dozens of others. It loops. It repeats. It wasn't made for you.

Ebbi is different. We built our own audio engine, called Reforge, specifically for this app. It powers every soundscape you hear. On iPhone, some sounds are generated live on your device from mathematical equations, with no audio files and no internet required; on Android, the same sounds are pre-rendered and stored in the app. Others are engineered from real-world acoustic analysis, reconstructed from their mathematical structure rather than copied from a recording.

Reforge isn't a feature we bolted on. It's the foundation the app is built on.

Researched. Engineered. Verified.

Three steps. Below, we say which sounds went through each one.

1. Researched

Every sound in Ebbi was chosen with published research on ADHD, cognition, and psychoacoustics in view, though the evidence is not equally strong for every sound. White noise has the strongest ADHD evidence: a meta-analysis of 13 studies of young people with ADHD or attention problems. Lo-fi rests on indirect evidence: children with and without ADHD made fewer errors with background music, though no study has tested lo-fi itself. Even silence is included because the research shows not every ADHD brain benefits from noise. Nothing is in the app without a reason, and every reason is documented with citations.

See the full research for every sound →

2. Engineered

For six of our sounds, Reforge analyzes real-world recordings and extracts their complete acoustic structure: every frequency, every amplitude change, every timing relationship. From that analysis, it builds a mathematical blueprint of the sound. That blueprint is what powers the audio you hear, not the original recording.

On iPhone, for core sounds like noise colors and rain, the math runs in real time on your device: your phone's processor generates every sample from equations. On Android, the same sounds are pre-rendered and stored in the app. Either way: no downloads, no streaming.

3. Verified

We tested our rain with audio classification technology: a machine learning model trained on a large collection of real-world recordings. We generated our rain, then asked the classifier: does this actually sound like what it's supposed to be? It recognized the sound as rain, far ahead of the sounds we did not want it mistaken for, such as wind, fire and static. That test ran on a rendered file of our rain design. The rain you hear is built from the same design, generated live on iPhone and as a newer render on Android, and we have not rerun the classifier on those versions.

We did not run the other sounds through the classifier. The noise colors and lo-fi are built from math alone, not modeled on a recording. The six sounds rebuilt from real recordings (ocean waves, fireplace, stream, wind, thunderstorm and cafe) are Reforge reconstructions, and we approved each one by listening before it shipped.

Why it matters for you

You probably don't care how the engine works. Here's what you get from it.

  • Sounds selected for your brain. Not a generic relaxation playlist. Each sound has its own research page with the evidence behind it.
  • Sounds that work offline. Every soundscape ships inside the app and plays offline. No internet, no streaming, no interruptions during a focus session.
  • Sounds we actually checked. Not just mixed by a sound designer and shipped. Our rain design passed an audio classifier test and was compared with a real rain recording, and we approved each of the six sounds rebuilt from recordings by listening before it shipped.
  • Built in-house. Not licensed. Reforge was developed by Forge Hut LLC specifically for Ebbi. It's not a third-party plugin or a stock audio library. It's ours, and it's patent pending.

Under the hood

For the curious. Not required reading.

The classifier we used is a machine learning model originally designed to identify sounds in the real world. We repurposed it as a quality gate for engineered audio: instead of asking "what is this sound?" we ask "does this generated sound match what it's supposed to be?" So far we have used it on our rain. This approach, using audio classification as a verification step for synthesized output, is covered by our pending patent application.

For rain, we also scored how closely a render of our rain design matched a real rain recording in timbre, spectral shape, how its loudness rises and falls, and brightness. It scored 0.815, above the 0.80 target we set.

For the research behind each individual sound, visit the Research page.

Patent Pending

Reforge is a trademark of The Forge Hut, LLC.

← Back to all research

Ready to focus differently?

Download Ebbi for iPhone or Android. Or join the focus-tips list.

Download on the App Store Get it on Google Play

Try Pro free for 14 days. Cancel anytime.