Can you share more information on the post-LLM processing and the prompt you use? I would like to try this out but don't see any post-LLM options in Handy.
replyedit: nevermind, found info on the docs about how to enable post processing. Would still be interested in your prompt though if you don't mind sharing!
You have to enable "Experimental Features" under "Advanced."
replyThis is the prompt I use (it's probably overkill and can be condensed):
What is Parakeet?
replyI believe this is the correct link. I use it too in Handy, for English and Spanish transcriptions:
https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3
replyMaybe they meant narakeet?
replyParakeet is the name of a speech to text model from Nvidia. Roughly comparable to whisper from openAI.
replyIt's the model doing the work inside the wrapper that an app provides.
Yep, here's the v2 and v3:
replyhttps://huggingface.co/nvidia/parakeet-tdt-0.6b-v2
https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3
It's almost instant on my new M5 Max w/ 36GB of memory, but I used both with Handy on my previous 2019 Intel Mac w/ 16GB memory and was completely surprised at just how fast it was for being on-device! Not instant, but only a couple seconds.
I’m using it on an M3 max 32gb, and I’m getting 60-70x realtime for recordings and crazy good accuracy. I can get an hour of audio transcribed in a minute. Similar results from Whisper, but half the speed.
replyTranscription this good used to cost A LOT, now it rounds down to free.