What is the best open-source options to create a pipeline like ElevenLab (Speech-to-text, brain LLM and text-to-speech)

Posted by frequiem11@reddit | LocalLLaMA | View on Reddit | 4 comments

I want to create a pipeline locally hosted and we can't use a outsource provider due to regulations. There are two ideas in my head. 1- Create a locally hosted pipeline, if so what are the best way to overcome this? 2- Find a way around to use ElevenLab (maybe redact sensitive data or some other techniques?)

4 Comments

[-]

rhinodevil@reddit

I built a little sample pipeline for this with my "wrapper" libraries based on Whisper.cpp, llama.cpp and Piper, in C/C++, here: [https://github.com/RhinoDevel/mt\_llm/tree/main/stt\_llm\_tts-pipeline-example](https://github.com/RhinoDevel/mt_llm/tree/main/stt_llm_tts-pipeline-example)

FullOf_Bad_Ideas@reddit

https://github.com/kyutai-labs/unmute

frequiem11@reddit (OP)

thanks , have you ever used this?

briefly, I had it running locally. Swapping LLMs was super easy. It's one of the best low latency fully local voice ai chat pipelines.

Reply to Post

4 Comments