Same thing I’ve been playing with these week. I want to carve out some time to take chunks from the streaming to create a loop of sending and receiving from the TTS API for super low latency I think the biggest problem is maintaining continuity of the voice so perhaps render the first sentence locally and while it plays assemble the next audio file or two server-side then send back?????
you can get faster calls using Groq new LARGE whisper, and Llama 3.1 70bn if you stream the audio ASAP and have tiny chunk sizes you can down to sub 1 second responses
@@zipaJopa I can dig one up and put it on git gimme a few hours, I have an iOS version but it’s faffy. Will reduce it to python and write a simple readme
@@zipaJopa Just rewrote the script will post it to GIT in a few mins and share the link here. (y) might make a video on it actually... Thanks for asking
Love the content but I think it would be helpful to at least mention relative costs. You've said in the past that the API costs are worth the investment. You are likely correct but probably still worth mentioning. Thanks again for the content.
"Personal AI is TOO valuable to leave in the hands of Big Tech"... proceeds to build his AI assistant with APIs hosted by "big tech". Kinda expected to see whisper, llama, mycroft or sth.
Your content is top notch man. It’s really easy to digest.
the #1 video I look forward to each week, sent you an email as well
That first one may as well have called you senpai. 😂
Amazing, so much value in a PA. Can't wait for speech-to-speech models to come to the market, super natural convos.
As we say in the UK - this is the dogs bollox! Fantastic work as always Dan.
would it be possible to further reduce latencies by streaming transcription and streaming response
That’s what I was thinking too
Amazing content - keep up this combo of practical + high level videos 💪
Same thing I’ve been playing with these week. I want to carve out some time to take chunks from the streaming to create a loop of sending and receiving from the TTS API for super low latency
I think the biggest problem is maintaining continuity of the voice so perhaps render the first sentence locally and while it plays assemble the next audio file or two server-side then send back?????
you can get faster calls using Groq new LARGE whisper, and Llama 3.1 70bn if you stream the audio ASAP and have tiny chunk sizes you can down to sub 1 second responses
Do you have any repos on hand that showcases this setup?
Many thanks!
@@zipaJopa I can dig one up and put it on git gimme a few hours, I have an iOS version but it’s faffy. Will reduce it to python and write a simple readme
@@zipaJopa Just rewrote the script will post it to GIT in a few mins and share the link here. (y) might make a video on it actually... Thanks for asking
@@6lack5ushi my hero! 💕
@@6lack5ushiwheres the link?
that's neat, I've been trying to get TTS to work for open-interpreter which I use a lot with gpt-4o-mini
do you know what the word "own" means?
What would be the main difference between the custom voice assistant and the OpenAI voice mode?
This was so great. I left you a PR on the repo, waiting for your feedback. Would love to demo it for you.
Is there a way to run this on your mobile device and have it in a call/open conversation function?
Amazing job !!!
Love the content but I think it would be helpful to at least mention relative costs. You've said in the past that the API costs are worth the investment. You are likely correct but probably still worth mentioning. Thanks again for the content.
Does anybody know a STT model that could be run locally for live transcript?
The top area for this will be Cybersecurity hands down.
In years time there will be entire cyberwars being waged and run by AI systems.
perfect
Eleven labs pricing structure is prohibitive. Very prohibitive.
"Personal AI is TOO valuable to leave in the hands of Big Tech"... proceeds to build his AI assistant with APIs hosted by "big tech". Kinda expected to see whisper, llama, mycroft or sth.
this is just a voiceflow type chatbot, it's not your own assistant.
Good point. I don’t really understand the point of this.
cringe lol