[Python] RuntimeError when transcribing audio using whisper.cpp with WebAssembly in browser

I'm trying to run whisper.cpp compiled to WebAssembly (whisper.wasm) locally in my browser on macOS. The goal is to record audio using the microphone, transcribe it using a base model, and get the results client-side.

Environment

• macOS
• Browser: (e.g., Chrome)
• Installed tools:

brew install emscripten
brew install libomp
export OpenMP_ROOT=$(brew --prefix)/opt/libomp

Steps to Reproduce


git clone https://github.com/ggml-org/whisper.cpp
cd whisper.cpp
mkdir build-em && cd build-em
emcmake cmake ..
make -j
cd bin
python3 -m http.server


Then I open http://localhost:8000/whisper.wasm/, and:

1. Download a base model (e.g., base.en).
2. Record audio via the microphone saying: “hello, testing testing”.
3. Click the “Transcribe” button.

Problem


At the transcription step, I get the following error in the browser console:

main.js:1 Uncaught RuntimeError: Aborted(). Build with -sASSERTIONS for more info.
at abort (main.js:1:6458)
at __abort_js (main.js:1:1363067)
at 003b63c6:0x77ca
at 003b63c6:0x415b
at 003b63c6:0xc8a30
at 003b63c6:0xc8b1a
at Object.full_default (eval at craftInvokerFunction (main.js:1:1), <anonymous>:9:10)
at whisper.wasm/:647:42

Question


What does this RuntimeError: Aborted() mean in the context of running whisper.wasm, and how can I debug or fix it?

Do I need to rebuild with specific flags (e.g., -sASSERTIONS) or is there a runtime/browser config I might be missing?

Any insight would be appreciated!

Continue reading...