apt install -y git-lfs
mkdir /piper-voices &&
cd /piper-voices &&
git init &&
git remote add origin https://huggingface.co/rhasspy/piper-voices &&
git sparse-checkout init --cone &&
git sparse-checkout set en &&
git config --local lfs.fetchinclude "en/**" &&
git pull origin main
Consider:
- Look at starting speakturbo-daemon behind the scenes
- Replacing speakturbo with Piper, and ability to clone voices locally.
Modelfile Params: https://github.com/ollama/ollama/blob/main/docs/modelfile.mdx
Not bad, but tends to over think and then suddenly gets really dumb. Need to play more, but it's kinda slow.
I thought the q4_K_M variant was a thinker, this one is insanely big thinking.
/no_think doesn't seem to have any effect , which is annoying as hell.
honestly, pretty trash. didn't really do anything i asked, even though it claimed it did.