Pocket-TTS seems to be a TTS server that returns audio much faster locally than Piper, so I built a container that enables it via Wyoming protocol and zeroconf to be available in Home Assistant.
There is the ability to use an audio snippet to clone a voice that would be used by pocket-TTS, I haven’t enabled anything like that yet. PRs welcome.
Thanks, I forgot to link that. I guess it’s in the README in any case.
English only at the moment. Too bad.
Good job on the Wyoming part of it!
Thanks. Yah, afaik pocket-tts hasn’t built anything but English yet. It’s pretty new.
I can’t pull the image:
docker run -d \
--name pocket-tts-wyoming \
-p 10201:10201 \
-e DEFAULT_VOICE=alba \
-v pocket-tts-hf-cache:/root/.cache/huggingface \
-v pocket-tts-cache:/root/.cache/pocket_tts \
pocket-tts-wyoming
Unable to find image 'pocket-tts-wyoming:latest' locally
docker: Error response from daemon: pull access denied for pocket-tts-wyoming, repository does not exist or may require 'docker login': denied: requested access to the resource is denied
I am logged into Docker as well via docker login.
edit I cloned the repo via Git and was able to get it to build and run.
Good to hear you got it to build. The base project seems to move fast so I just set this up to require a local build which will pull the pocket-TTS current code in directly.
I guess I could post a GHCR image, but I’ve never done that. I just build my own images, it’s pretty quick and then they’re up to date.
For me personally, I’d rather pull from a GHCR image than build from the source, I think too, this would garner more users who don’t want to or cant build it for one reason or another.