#Voiceassistant
1 messages · Page 1 of 1 (latest)
The Voice PE uses microwakeword, which means the wakeword detection is done on the device itself. Microwakeword doess't support custom wakewords
It actually does.
It's just harder to train those models, and put them to the YAML takes some time. But it's 100% possible.
However, models for MWW require fine-tuning, and without real voice samples they're not that great.
Will it be complicated to embed your trigger word? "Homie," what do I need to do to do that?
You need to trail the model first. For that you need PC with Nvidia card.
Also general best practice is a wake word with at least 3 syllables, 4 or more is even better.
Pastebin.com is the number one paste tool since 2002. Pastebin is a website where you can store text online for a set period of time.
You can find some online pre trained
Your config is pointing to local JSON. /config/esphome/micro_wake_word/hey_glados.json
Correct, I put the glados wake word there
I have this thing with disappearing, gitrepros
I mean, maybe it's time to share models.
A quick web search will find many
It’s how I lazy found mine
I’ll get around to having a llm write a nice read me in a git repro somewhere once I get some major changes done on my homelab
There's a lot for OWW, not that much for MWW.
I think there’s guides online and containers to train them for mww I think that’s where this came from
It never goes to that. There's always something to change on home lab ;)))
I know. I was posting one of repos with colab. But there's some hardware and knowledge involved to train one.
I think there’s a repro out there with an entire tookit
https://github.com/SmolinskiP/GLaSSIST/tree/main/models here are onnx and tflite, but can't find whre are json for tflite
Ah it’s on a ha forum post
Let me find it again
I am searching for alexa.onnx that is the same that ESP have, since its international and works well
So i trained my own micro wake words and i wanted to share them with you all add this yaml to your config and it should give you the wake words “GlaDOS” “Hey GlaDOS” “GlaDOS Please” this is a refrence to the GlaDOS robot character in portal. I combined it with a trained glados assistant voice and llm with a glados prefix to the l...
You might find a working Alexa on there
I was able to successfully build a model for ‘computer’ using the MasterPhooey Docker Trainer + jupyter notebook advanced model. I ran it on Win11 using DockerDesktop and the suggested run commands. It took maybe 15-20 minutes to run the notebook but it worked. Then I forked the repo at GitHub - JohnnyPrimus/Custom_V2_MicroWakeWords and cre...
Oh that’s the micro wake version :p
yep a lot of mww not many oww
only one that I managed to find is at:
https://github.com/msp1974/ViewAssistCompanionApp/tree/dev/app/src/main/assets
but its American version, so not really doing like ESP mww does
Ah
What are you using open wake word on?
maybe that is what I am after for, thanks, will test it with GLaSSIST now
VACA is my primary target, waiting for developer to implement downloading of oww
works great so far for android tablets and phones
Ohhh that’s cool
works locally on android device as wake word
you will need this as well:
https://github.com/dinki/View-Assist
it seems that you saved me @empty saddle seems that alexa_v0.1.onnx is the one that I searched for, thanks 🙂
Gold mine for you
https://docs.openvino.ai/2024/openvino-workflow/model-server/ovms_demos_continuous_batching.html
oh. I wonder how much faster whisper runs on an intel NUC
NVIDIA NeMo Parakeet-TDT is crazy good on intel n100
have you seen the new go2rtc release? wyoming support to have cameras become voice satellites 😄
Is there another way? Back when I had a prototype assistant, I used the open wake word, and it worked somewhat. I created the word HOMIE back then, and it even worked decently.