Create a Modal.com FastAPI script that will take an input vi...

Prompt

Create a Modal.com FastAPI script that will take an input video, transcribe the speech from the entire video with timestamps with a separate transcript per speaker, and store the transcripts and the title of the video file they came from, and create a searchable index by word or phrase of every transcript, that's updated after each video transcription. The index will return the surrounding text, video name, and timestamp. It should have a relatively simple API spec. Do not use any gated repositories for open-source models. For the Modal image, make sure to use Ubuntu due to the recent archival of Debian dependency repositories.