Well, that's why it's free and code is available, so everyone can improve it. It was a fun weekend project to test the ability of making an endless stories by various open source LLM models. Also couldn't find any good sounding real time TTS model, hopefully someone will drop some suggestions.
@@zyxwvutsrqponmlkh Yes, have a look on the github link. If you have any ideas, I'm happy to listen. Code can be changed also for bigger context window, so the stories will have more sense. A lot can be improved and it could be a good start project for someone. There is also many options for style, so you can have some crime, documentary or some funny senseless talking, everything done in realtime. Have a try and experiment a bit, it's actually pretty funny to listen LLMs endless hallucinations. You need nvidia with 8GB VRAM at least though. By default script downloads uncensored LLama3.1:8B model, but you can try any other that is Ollama friendly