Post AcMu9dX9JebLQPaGO0 by ranfdev@linuxrocks.online
(DIR) More posts by ranfdev@linuxrocks.online
(DIR) Post #AcMu9cEg8qETOpvzxQ by ranfdev@linuxrocks.online
2023-12-01T14:30:24Z
1 likes, 1 repeats
I think my next project will be a little voice assistant running locally. Now that we have llamafile https://github.com/Mozilla-Ocho/llamafile running an AI locally has become simpler than ever.I just need to write a program which listens for a keyword like "Hey Google" in the background, then transcribes the spoken text with whisper.cpp and then feeds the transcription to llamafile.Finally, I need to find a decent open source TTS capable of running on the CPU in realtime.Then I build a GTK UI to show the current conversation, and I package everything as a flatpak.
(DIR) Post #AcMu9dX9JebLQPaGO0 by ranfdev@linuxrocks.online
2023-12-01T14:34:10Z
0 likes, 0 repeats
Since the AI is going to run locally, it can read your clipboard without any privacy issues. I plan on using LLaVa as the underlying model, because it's multimodal, so that you could copy an image and then ask "Hey Lava, can you describe the image in the clipboard?".