Migrate space to Fal.ai endpoint

As an alternative to duplicating Gradio Spaces (which adds overhead and overcosts) or hacking them to add real-time (which they are not designed for - same for Streamlit, it has real-time capabilities but is not designed to run as an API so we can't use it) I think it would make sense to create private endpoints on Fal.ai - https://fal.ai/docs/migration-guidelines/replicate - https://fal.ai/docs/private-serverless-models/real-time-endpoints