Joey D’Antoni explains why you almost never want to use FILESTREAM:
Yesterday, I accidentally walked into a discussion on LinkedIn about the merits of Filestream in SQL Server. If you aren’t familar with Filestream, consider yourself lucky, it was a feature that was added to SQL Server in the short timeframe that people thought it was a good idea to use databases for file storage, and before the enlightened times when object storage became a thing. I first remember blobs being a thing in Oracle 8i, where at least you had the ability to store them in a tablespace with a larger block size than 8k, and had a dedicated area of the buffer pool with that larger block size that you could dedicate to that blob tablespace.
Joey has the right of things. There are rare exceptions where it could make sense to store files in databases. My best example involves storing ML models that we use in SQL Server ML Services, simply because of how difficult it is to read anything off of disk via ML Services. But that’s a real edge case.