wrong websocket address

[manuliner]

to make this work i needed to change the websocket address in this line

return Storage.Local.GetOrAdd("Host", "wss://speech.platform.bing.com");

to

return Storage.Local.GetOrAdd("Host", "wss://{your Deployment ID}.api.cris.ai");

i also changed this in my gulped JS file under node_modules/microsoft-speech-browser-sdk/distrib/src/sdk/speech.browser/SpeechConnectionFactory.js

you should change this in your sample, as it is not working at the moment

[zhouwangzw]

The sample is for using Speech Recognition API(using general-purpose model without customization), so the endpoint "wss://speech.platform.bing.com" is correct.
If you want to use Custom Speech service with your own customized model, you need to use the endpoint with your own deployment.

Thanks,

[manuliner]

hello @zhouwangzw,
Custom Speech Service DOCs tells me to use these samples.

Subsection OPUS
https://docs.microsoft.com/de-de/azure/cognitive-services/speech/api-reference-rest/websocketprotocol

[zhouwangzw]

On the Custom Speech Service Doc, "Send requests by using the Speech Protocol" asks you to replace the endpoint in the sample with your own deployment endpoint:

Currently, the only official client implementation is for JavaScript. If you want to start with the sample provided there, make the following changes to the code and build the sample again:
In src\sdk\speech.browser\SpeechConnectionFactory.ts, replace the host name "wss://speech.platform.bing.com" with the host name shown on the details page of your deployment. Do not insert the full URI here but just the wss protocol scheme and the host name.

The OPUS section you mentioned is for the general-purpose speech recognition API (instead of custom speech API), so it uses the "wss://speech.platform.bing.com" endpoint.

[manuliner]

maybe u can help me on my next issue:

my 15 seconds test file is working, but when i change to my 3 minutes testfile, cutsom speech service is not recognizing the whole file(stops after 15 sec). Any idea what i can change?

i tried recognitionmode dictation and conversation
implemention as in the sample

[zhouwangzw]

For custom speech issues, please contact the custom speech service team. We do not own their service.

[manuliner]

how to contact?
all contact possiblities forward to this git repo

[solidcloudio]

Our services aren't available right now

We're working to restore all services as soon as possible. Please check back soon.

Ref A: E46EE7F8DE28480E89BA1F908D77D2DC Ref B: PHX01EDGE0217 Ref C: 2018-03-29T15:44:18Z

[mikebranstein]

See pull request #82 that addresses this issue.