I get this error quite often
home assistant speech-to-text failed (stt-stream-failed)
and in the whisper add on logs I see:
DEBUG:__main__:Namespace(model='auto', uri='tcp://0.0.0.0:10300', data_dir=['/data'], download_dir='/data', device='cpu', language='en', compute_type='default', beam_size=0, initial_prompt='null', debug=True, log_format='%(levelname)s:%(name)s:%(message)s')
DEBUG:__main__:Model automatically selected: tiny-int8
DEBUG:__main__:Beam size automatically selected: 1
DEBUG:__main__:Loading rhasspy/faster-whisper-tiny-int8
DEBUG:urllib3.connectionpool:Starting new HTTPS connection (1): huggingface.co:443
DEBUG:urllib3.connectionpool:https://huggingface.co:443 "GET /api/models/rhasspy/faster-whisper-tiny-int8/revision/main HTTP/11" 200 719
DEBUG:urllib3.connectionpool:https://huggingface.co:443 "HEAD /openai/whisper-tiny/resolve/main/tokenizer.json HTTP/11" 200 0
My current theory is that this is because of my duck DNS remote access setup, where I wasn’t able to forward all the ports I wanted to.
My FritzBox router allowed me to forward the 8123 port but not the 443 port as well.
Is this plausible? Do I need to either get a router that will do support both ports being forwarded, or switch to Home Assistant Cloud?
Just got a HA Voice: Preview Edition and its great when it works the first time, but then I have to restart whisper to get past that sst stream failed error…
any theories welcome.