pipecat-ai
Kokoro, Orpheus, and CSM Support
How do I make it work with TTS like Kokoro, Orpheus, or CSM? Generally, how do I make it work with custom TTS?
Pipecat doesn't support those TTS services, so you'd have to write a service implementation for them. This can be something you either do for your own product/project or if you'd like, you can submit them to the project for us to review.
I already have added Moonshine ASR and Kokoro to pipecat for fully local speech to speech here. Should I raise a PR for both?
I can also add deepinfra for TTS Orpheus, and CSM.
@marmikcfc we're happy to review those contributions, if you're interested in sharing with the community 🙇
I already have added Moonshine ASR and Kokoro to pipecat for fully local speech to speech here. Should I raise a PR for both?
I can also add deepinfra for TTS Orpheus, and CSM.
Hi @marmikcfc - I integrated your kokoro and moonshine modules in my fork (and added attribution to you in the modules so people won't think it's my work) along with a couple new examples. Would it be okay for me to contribute those back into main? I ask because it's primarily your code and don't want to step on toes.
Hi @CleverLittleMaker , thanks I've been meaning to take a recent pull but have been busy with work and test and merge it again. If you've already done it feel free to raise a PR. :)
Guys, thanks for all your hard work, (Pipecat team and the community) Orpheus TTS support (or instructions of implementing that) would be greatly appreciated.
Is this issue still open, I have added the kokoro (tts) for the fastapi websocket, so that you can use it for the deployment purpose.
@markbackman I can help in integrating the kokoro with the Fastapi websocket, let me know how i can contribute to this new service.
in the services addition i am able to write the code for the sarvam speech to text(STT) and it is working fine how i can contribute that in your codebase @markbackman
@Jaiaggarwaaaaal can you please share the source code of kokoro with FastAPI
@murugancmi ok i will create the pr for that, you can take the help from there