AudioCraft

Problem

Music students record long lessons but practice in short sections, and phones had no good way to split, name and loop those sections.

What we did

We designed and shipped an app built around segments: many named practice sections from one recording, stored without duplicating audio, plus merging tracks with per-track pitch and speed.

Result

5,000+ downloads and 650+ monthly users with no marketing spend. And one clear lesson: niche products need niche positioning.

Where it started

Carnatic music is South Indian classical music, learned largely by ear from a teacher, so most students record their lessons, often 30–60 minutes long, and practice from them.

But practice doesn’t work like listening. You don’t play a 45-minute recording from start to finish. You find the two minutes where your teacher demonstrated a phrase, loop it, and drill it until it sticks. On a desktop that’s easy. On a phone, where practice actually happens, it was painful.

What existed

We reviewed the top audio editing apps on Google Play. They solved the same jobs (trim, merge, convert), and the two features closest to what learners need both fell short.

Nothing let a learner turn one long recording into a set of named practice sections and play through them without touching the phone.

Key decisions

01 Make segments the core of the app

Options

Build another audio editor with a loop feature, or build the app around segments.

We chose

Segments first. A recording becomes a playlist of practice sections, with editing tools around it.

Why

That was the job no existing app did well, and the reason a learner would switch.

Trade-off

A narrower story to tell. That tension came back when we positioned the app.

02 Store segments as references, not new files

Options

Export each segment as its own audio file, or save only its start and end points.

We chose

Save start and end points as metadata in “work files” that point to the original recording.

Why

A 45-minute lesson is about 40 MB. Ten exported segments would use ten times the storage for the same audio. As references, users get unlimited segments at almost no cost in space.

Trade-off

More engineering work, and it needed close collaboration with the developer to get right.

03 Don’t use AI to find segments automatically

Options

Have AI detect and suggest segments, or let learners mark their own.

We chose

Learners mark their own.

Why

Which sections matter depends entirely on the learner. One student needs to drill the opening phrase; another needs the closing variation. It’s a workflow problem, not a detection problem.

Trade-off

Learners do a little manual work for each recording.

04 Replace “A–B” with plain words

What we saw

Our first design used the “A–B” labels from music software. We tested it with about 25 musicians and 15 general users. Musicians understood it immediately; general users didn’t know what A and B meant.

We changed

Segment creation became “mark start, mark end, name it”, with the A–B labels removed entirely.

Result

Task completion improved in both groups, and the app became usable for people who aren’t trained musicians.

We spent nothing on marketing and relied on App Store optimization and organic search.

To get more downloads, we positioned AudioCraft as a general “audio editor” rather than a “music practice tool”. The logic was sound on paper: more search keywords, more visibility, more installs. It worked, but it attracted the wrong people.

• Downloads grew, but most new users wanted basic trimming and merging.
• Segmentation, the feature that made AudioCraft different, was buried behind generic branding.
• The musicians it was built for couldn’t easily find out it solved their exact problem.

We optimized for volume over fit.

About one in eight people who ever installed AudioCraft still use it every month, with no paid acquisition behind it.

What the numbers don’t tell us

We can’t yet see whether the same people return each month or whether new users replace those who leave, how many people use segmentation versus basic editing, or whether our regular users are the musicians we built for. We suspect they are, but we haven’t confirmed it.

What worked

• The core insight was right: music learners need better mobile tools for practice.
• Storing segments as references kept the app light and made unlimited segments possible.
• Dropping the A–B labels opened the app to people without music training.

What didn’t

• We built for a niche and marketed to everyone, so downloads rose but fit didn’t.
• We didn’t add analytics in the first version, so we can’t answer basic questions about how segmentation is used.
• We haven’t tested whether people would pay, and there isn’t enough volume for ads to matter.

We’re repositioning AudioCraft as a music practice tool with editing built in, instead of an editor with practice features. Fewer people may download it, but the right people should find it, stay, and be more likely to pay. If the repositioning works, splitting it into a free editor and a paid practice app is the next option, though that means two apps to maintain.

The first step is small: rewrite the store listing around practice and measure whether new users stick around better.