Hello! I am JM, the creator of TalkPersona. This experimental project is aimed at developing a unique AI companion with real-time video and natural voice capabilities, that you won't see in any other public spaces.
This artificial intelligence chatbot is computationally intensive, as it includes quite a few neural networks, including an LLM (large language model) which is uncensored (so it won't be boringly polite or overly politically correct like other public chatbots) and is self-run (not relying on external APIs), transcription, text to speech, wave file pattern extraction, and face animation with lip sync. There might be some initial latency when the AI chatbot starts talking, but it should smooth out after two seconds. You have the option to disable video to talk to the AI with audio-only if you want an even faster response. The speed experience may differ depending on where you are located.
As it requires quite some compute, I'm only running one chatbot at selected times for the moment. It can be scaled up to multiple chatbots if there is demand. There are some future enhancements planned, including different voices, different images to generate the face animation, giving user the option to upload a voice wav file for the chatbot to clone it, saving of conversation history for next day, and so on. Feel free to leave me a feedback.
You can reach me by emailing me at: [email protected]