Memory issue when model can not load.

[MarketingPip]

The library uses excessive memory when the model is NOT found in path provided.

As well blobs should be provided etc. Tho could be turned into base64 url?

[thurti]

The library uses excessive memory when the model is NOT found in path provided.

I will check the init method. I guess the error handling needs some improvement.

As well blobs should be provided etc. Tho could be turned into base64 url?

Can you please explain this in more detail? Where should a blob of what be provided?

Regarding the model file, the model option can be of type File | string . You can use the File API to create a new File object from a blob or a base64 encoded string.

[MarketingPip]

@thurti I was thinking passing blob for transcribing - tho that might be stupid (since we can just turn blob into base64 string right? (so maybe just skip that?)

// transcribe audio/video file
const result = await transcriber.transcribe(BLOB here);

console.log(result);

but as for fetching / memory issue when model is not found via fetch etc.. Definitely needs a fix. For some reason I think the model is still trying to be init with nothing in it. (tho I have not looked in the code to confirm this)

[thurti]

I was thinking passing blob for transcribing

You can use the File API to create a file object from a blob

// transcribe audio/video file
const result = await transcriber.transcribe( new File([BLOB here], "myBlob.wav") );

but as for fetching / memory issue when model is not found via fetch etc.. Definitely needs a fix.

#11 will fix this in the next release.