[llama_person]

If you have access to a GPU, try a vision-language model using Ollama, and feed it your notes. Might work out!