msqr1/vosklet

A speech recognizer that can run on the browser, inspired by vosk-browser

Read vertically

Hi, I stuff. the

I like I most.

am to love

Rylex. code C++

vosklet's Issues

Not work with different model

On the step of model's creation (malloc) JS throw error without text but with 1 number, it looks like number of bytes tried to allocate.

Have you built model with fixed memory size? doesn't it allow memory_grow?

evaluation give worse metrics then server-based VOSK. Way to run it sync for test?

I have a dataset of syllables on which I perform the evaluation.
Each wav = 1 syllable (0.2 sec). Just read one by one, recognize, check with ground true.

Vosk on the host gives ~80% accuracy.

Vosklet give ~65%, Possible it's because async nature of the js wrapper.
I use a timeout 2 sec to divide different wav recognition, but 11% of records have nothing as recognized result, and 8% contains 2 syllables.

Other things all the same between hosted Vosk and browser Vosklet.
Maybe running in main thread synchronously will help?

Unhandled Promise Rejection: ReferenceError: Can't find variable: SharedArrayBuffer

[Error] Unhandled Promise Rejection: ReferenceError: Can't find variable: SharedArrayBuffer
	(anonymous function) (Vosklet.js:9:9494)
	(anonymous function) (index.html:21)

Safari Version 17.3.1 (19617.2.4.11.12)

Happens after clicking start and speaking via the example

Thanks for the lib, still failing to get browser offline speech recognition working with any example from any offering 🙃

working example?

I tried to run on IDE devserver, different services but get error about SharedBufferArray
I created a simple Flask with necessary CORS headers, but now it fails at

let model = await module.createModel("en-model.tgz","model","ID")

with

Uncaught (in promise) TypeError: Failed to fetch